Sybase数据库技术，数据库恢复专家

combined the CONTAINS predicate and an 'Into #temp' clause.

This happens in version 12.4.3 ESD7.

Exception Thrown from hqm_con.cxx:102, Err# 1, tid 264 origtid 264

O/S Err#: 0, ErrID: 8192 (hqm_exception)

CtLibrary Error: 141, Severity: 0, Origin: 1, Layer: 1

Error Message: ct_con_props(SET,PASSWORD): user api layer: external error: buflen must be 0 or CS_UNUSED when buffer is NULL.

OS Error: 0, OS Message: .

-- (hqm_con.cxx 102)

Probelm reason:

The INSERT... LOCATION statement always takes the user and password from the

current connection to connect to an ASE server. The catalog stores the password

only as hash value created by an one-way algorithm for security reason.

So the password string is not available for the event execution before the end user

using this username has connected to ASIQ.

As soon as an user has connected using the username all subsequent INSERT... LOCATION

in events work file

Its not possible to fix this problem without added new syntax or an new feature.

i'm going to create an Enhancement Request for this.

1-3 hours) the IQ server gets into one of the following situations:

- the IQ server crashes with the below stacktrace

- the IQ server prints the below stacktrace but still exist as unix process and

consumes about 380 percent CPU usage on a 4 CPU machine

The stack trace is:

ucstkgentrace__FilT2 + 0x70 [/sybase/t/sys/asiq12/bin//../lib/libiq.sl]

hos_ABORT__FPCcUiT1 + 0x1c4 [/sybase/t/sys/asiq12/bin//../lib/libiq.sl]

SigHndlr__Fi + 0x48 [/sybase/t/sys/asiq12/bin//../lib/libiq.sl]

_sigreturn [/usr/lib/pa20_64/libc.2]

_memmove + 0x22c [/usr/lib/pa20_64/libc.2]

mallinfo + 0x1f70 [/usr/lib/pa20_64/libc.2]

_ltostr + 0xb6c [/usr/lib/pa20_64/libc.2]

realloc + 0x1fc [/usr/lib/pa20_64/libc.2]

ut_do_realloc + 0x18 [/sybase/t/sys/asiq12/lib/libdbserv7.sl.1]

ut_realloc + 0x64 [/sybase/t/sys/asiq12/lib/libdbserv7.sl.1]

MsgListAddToRow__FPciT2 + 0xb4 [/sybase/t/sys/asiq12/lib/libdbserv7.sl.1]

This should be corrected. With Ansi_update_constraints = 'strict' an delete statement

returns: "update operation attempted on non-updatabled"

shows wrong data.

In IQ, we could select rows for login & pwd data out of the sysuserperm system table, but IQ would not allow the user to add (load/insert) rows to the system tables. So right now, we don't have the functionality in ASIQ to accomplish this though this could be done through scripts by the user.

database all characters are converted to ROMAN8 but the ASIQ db is ISO_1.

To prevent this convertion the LC_ALL variable on OS level has to be set.

With the request the connection done by the INSERT..LOCATION should always set the client local info to the same as the IQ db's collation sequence.

DBSPAWN ERROR: -81

Invalid database server command line

instead of causing server failure.

INSERT INTO <temp table> ... SELECT ...

did not generate a pre-run query plan as they normally would.

INSERT SELECT statements may be returned with an insufficient stack error

message requesting that you increase the -iqtss setting. Increasing -iqtss by 72KB permits such statements to complete.

On 32-bit systems only, depending on your memory configuration, you may need to decrease

Load_Memory_MB in order to increase -iqtss.

server failure. This problem is corrected.

INSERT INTO testtable VALUES (getdate())

This problem affected a query with a SELECT DISTINCT column in a derived table that was

not used in another query block. The column also happened to be in a case expression that

was an argument of a complex expression.

error and create a core file if an attempt is made to generate

an HTML file.

table is empty, have either a 1-byte or 2-byte FP index, and has neither a LF nor an HG index. Such a column, used in a predicate, could potentially cause either a floating point hardware exception and subsequent server failure or an "invalid arithmetic construction" error.

suspended during these scheduling conflicts.

field."

SELECT district, grade, avg(math_score),

rank() over (PARTITION BY district ORDER BY AVG(math_score)

desc) AS rank_d

FROM math_report

GROUP BY district, grade

ORDER BY district;

Such statements now return an appropriate error.

(char(9) and char(10)) returned the error "Non-Fatal IQ Internal Error Detected

at dfjo_Hash.cxx:378 on thread 273" followed by a stack trace.

using a HG index returned the wrong result.

For example:

SELECT col, COUNT(*) FROM T GROUP BY col

over a set of simple base tables has been extended to include

cases where the exact distinct count for the join column from the view

is not available, but the exact distinct counts for all the underlying

base columns are available. In these cases the cost analysis of the

pushdown join can proceed using the worst case collective distinct count.

column returned incorrect results.

For example:

SELECT * FROM test1 WHERE

(a > 1 AND a <3) OR

(a > 4 AND a <6) OR

(a > 8 AND a <10)

returned values 2,5,6,7,8 and 9.

a server failure occurred at st_server.cxx; 1037.

returned if the procedure has an IN parameter of type NUMERIC:

DIAG [HY104] [DataDirect][ODBC Sybase Wire Protocol driver]Invalid precision value. Error in parameter 1. (0) (Duplicate of CR 323070.)

with the error "correlation name A not found."

In 12.5 esd2, the error is "column '' not found." Duplicate of CR 324035.

an ASIQ server may fail while creating

a very large database with the default

main cache size.

IQ Internal error. Please report this to Sybase IQ support.

-- (dfe_HostVar.cxx 250)

Duplicate of CR 326228.

The problem does not exist in DBISQL (java)

*** SybaseIQ Abort at st_server.cxx:1037 for PID: 174 ***

connection to the remote server (from the SA engine) is not closed correctly when the

client disconnects. Connections from dbisql java are handled

correctly with the connection closed on disconnect.

Signal 10 error:

*** SybaseIQ Abort at st_server.cxx:1037 for PID: 5410 ***

Same as CR 324390.

*** SybaseIQ Abort at st_server.cxx:1037 for PID: 15979 ***

Caught signal 10, program abort

UPDATE CONCORD

SET A.err_cd =

(case when substr(A.packet,1,10) = 'HOST ERROR' then 'HOST ERROR'

when substr(A.packet,1,3) = 'TXF' then substr(A.packet,56,10)

else '' end),

A.err_gb =

(case when substr(A.packet,1,10) = 'HOST ERROR' or

substr(A.packet,1,3) = 'TXF' then 'E'

else 'O' end)

FROM CONCORD A

WHERE trdate like '20030625%'

occurs between 'backup' and any other connection performing

the first operation of a new transaction.

than the default value, it should appear in the query plan.

a value of '0001-01-01 00:00:00' into a DTTM index caused

server failure. Duplicate of CR 327307.

value greater than 10000 loads zero rows instead of exactly

one row.

To avoid the problem, always set BLOCK FACTOR less than or equal

to 10000 in the LOAD statement.

on a Sun platform and caused a server on HP-UX to

fail with a Signal 10 error:

SELECT ... SUM(b2.num1) AS num1

FROM (SELECT ...) b1,

(SELECT ... sum(1) AS num1

...

GROUP BY char12, date1) b2

...

GROUP BY char12, date1

Currently, the first parameter must be a date, time, or timestamp and cannot be a string. It is documented in the IQ Reference Manual that the first parameter (datetime-expression) to the DATEFORMAT function MUST NOT be a string.

While in ASA and ASA documentation, there is no such limit.

with one or more descending direction indicators (DESC),

the query was rejected with an internal error because of

incorrect propogation of the ordering direction information

within the query optimizer.

User A creates a table and before inserting data to the table creates an index for every column. While User A is creating the indexes, other users execute SELECTs on the table created by User A.

For example, when the option is set to 12, these numbers would compare as equal:

1.23456789012345

1.23456789012389

When the option is set to 13, they would compare as unequal.

Because doubles are stored in binary (base 2) instead of decimal (base 10), this setting gives the approximate number of significant decimal digits used.

This option affects equality tests between two complex arithmetic expressions, not those done by the indexes.

Server restart is not required for the option to take effect.

In the 12.5.1 release the extention will be ".srvlog" on all platforms.

was rejected at dfe_ArithmeticExpression::209

(ASA Error -1001006: Data exception -

data type conversion is not possible.)

This query failed for m1 of datatype INT

or NUMERIC on both Sun and NT.

that has a primary key and/or unique constraint and

- IGNORE CONSTRAINT UNIQUE was specified, AND

- one or more UNIQUE constraint violations occurred

then it was possible that the primary key and/or

unique constraint would have been incorrectly built.

The user would have noticed this either by a query

returning wrong results or sp_iqcheckdb reporting

that a HG index used to represent the primary key

or unique constraint was inconsistent and missing

various keys and row ids.

Container Statistics

=======================================

Database Objects Checked','485012'

B-Array Count','4643'

.

.

'Bitmap Count','468536'

***Bitmap Verify Errors','2','*****

The IQ Message File will also contain two errors for each 1 bitmap problem it finds. The Message File will contain two failures that look similar to:

2003-09-20 05:20:20 0000000007 Exception Thrown from s_bmn.cxx:91, Err# 1, tid 231 origtid 231

and

2003-09-20 05:20:20 0000000007 Exception Thrown from s_bmn.cxx:100, Err# 1, tid 231 origtid 231

DATEPART(<part-name-constant>, <date_and_or_time_col>) = <constant>

would in the presence of a DATE, DTTM, or TIME index fail to use that

index to compute a precise selectivity estimate.

unexpected results depending on datatype due to

an incorrect condition check.

Wishes to be able to rename the IQMSG file from IQ, or with an external utility like DBLOG instead.

-debug flag, the output lines that began with "CatCountLock" and "CatUserLock"

included incorrect statistics for "Locks", "Waits", and "Wait%".

could result in server failure with the error "Fatal error:

memory exhausted" reported in the IQ server log.

would cause server failure in 12.5 but was ignored in 12.4.3.

Now an error will occur if FILLER() is specified.

FILLER(1) may be used instead of FILLER().

for Referential Integrity (primary/foreign key relationships) increases.

(moderate being defined for an N CPU system as between N+1 users

and 2*N users) the query engine would choose to go more parallel

then was optimal for a server that heavily loaded.

Msg 102, Level 15, State 0:

ASA Error -131: Syntax error near '(end of line)'

All subsequent statement will go fine

expression where the result of the expression can be

determined at query optimization time are now simplified

by the query optimizer.

This problem could cause the sp_iqcheckdb stored procedure to stop responding.

The system would appear idle (no CPU/IO activity). A stack trace of the running

server contained a frame with the name SetBitsGetDuplicates.

The check box has been removed because IQ databases do not currently support data encryption.

was running, the server could fail with no error message.

multiple user connections.

containing all of the following:

(1) an uncorrelated IN subquery, for which the optimizer

had selected the sort-based execution algorithm

(2) at least one additional subquery predicate

(3) both subquery predicates involved columns from

a single large table

(4) that large table was joined to a smaller table

(5) the optimizer had selected the NLPD join algorithm

to execute that join

would occur when opening a TCP/IP connection.

**Resolution**

Not targeted for any future release of ASIQ

return inconsistent results or cause server failure:

1) There was a simple predicate on a column

2) That column contained one or more NULL values

3) The column was being compared to one or more

constants that were beyond the dynamic range

of the column's data type

4) The effect of the predicate should have been to

pass all non-NULL values in the column

For example:

T.COL_TINYINT < 300.

The specific symptoms of this problem depended

on the precise combination of data and query.

Duplicate of CR 351513.

CONTAINS in the WHERE clause would ignore the CONTAINS criteria.

You have run out of IQ STORE dbspace in database D:\Sybase\ASIQ-12_5\demo\asiqdemo.db. In another session, please issue a CREATE DBSPACE ... IQ STORE command and add a dbspace of at least 8 MB.

create table t1 (c1 int, c2 int, c3 int)

insert t1(c3) values (9)

create hg index hg1 on t1(c1, c2)

sp_iqcheckdb 'verify index hg1'

Reaching the 1024 open file descriptor limit could result in

"too many files" or connection errors and could cause INSERT/LOCATION

to fail with "Connectivity libraries cannot be found".

start_asiq -n my_db_server -x 'tcpip{port=7934}'

-gd dba -iqdroplks my_db /work/database/my_db.db

This information was omitted from "Repairing allocation problems" in Chapter 2, Sybase IQ Troubleshooting and Error Messages Guide.

on columns having datatype 'long varchar' or 'long binary'.

The reason for this request is that certain types

of data (e.g., jpeg) do not compress. So by allowing

users to disable compression, the cost of trying

to compress data that won't compress is avoided thereby

saving CPU cycles.

expressions or functions over columns from the NULL

generating side of the outer join could be rejected

with an internal error. If this error is hit, the

stack trace will contain "dfo_Join::NeedCloneDFE".

Start ALLOCATION table: nation

Start ALLOCATION processing index: ASIQ_IDX_T445_C1_FP

Start ALLOCATION processing index: nationhg1

Done ALLOCATION table: nation

Done ALLOCATION processing index: nationhg1

Done ALLOCATION processing index: ASIQ_IDX_T445_C1_FP

sp_iqcheckdb ' mode target' (Where 'mode' is as it was before.)

target:[indextype idxtyp] database | database resetclocks | { [indextype idxtyp] table table-name | index index-name [...] }

idxtyp: is one of FP, CMP, LD, LF, HG, HNG, WD, DATE, TIME, DTTM

Example: To CHECK all of the FP indexes in database asiqsemo...

sp_iqcheckdb 'check indextype fp database'

Example: To VERIFY the FP and HG indexes in table customer and the lf indexes in table tableau...

sp_iqcheckdb 'verify indextype fp indextype hg table customer indextype lf table tableau'

If the user specifies an indextype that does not exist in the target, error "The specified indextype does not exist in the given target. -1000219" will result. For example, if there are no indexes of type "date" in table "blah", then

sp_iqcheckdb 'check indextype date table blah' will result in this error. If there are some hg indexes in the table, then

sp_iqcheckdb 'check indextype date indextype hg table blah' will check only the hg indexes.

when the following conditions were true:

a) it had a table (already loaded) with a date or datetime

column, and number of distinct year values in the DATE or DTTM index

exceeded 200

b) the user performed an incremental insert into the table

expressions containing a varying host variable

plus constants or additional host variables,

but containing no columns or aggregates.

from a very large table, several minutes could pass

between checks for whether the user had requested

that the statement be cancelled or the connection

dropped. Additional cancellation checks were added

to make the maximum delay during this type of

operation be a few seconds.

IQ Server should return an error message correctly.

INSERT [ INTO ] owner.tablename [ ( column-list ) ]

insert-load-options

[ LOCATION 'servername.dbname' [ ENCRYPTED PASSWORD ] [ PACKETSIZE packetsize ] ]

{ select-statement }

If "ENCRYPTED PASSWORD" is specified and the remote server does not support Open Client Library default password encryption, an error will result indicating an invalid userid or password was used. Sybase IQ, when used as the remote server, does not support this password encryption.

(-end of update-)

Password encryption always on when connecting to OpenClient with INSERT.... LOCATION.

10 aug 2004 update: Default Open Client Library password encryption has been enabled for INSERT...LOCATION when IQ connects to ASE. This new functionality has been added for the 12.6 release.

First, create the table to hold the user information and insert a row for each user.

create table dba.event_table ( username char(128) not null,

failed_login_attempts integer,

primary key (username) );

Create the event handler that will increment the number of failed login attempts.

create event ev_badlogin type ConnectFailed handler

begin

declare uid char(128);

declare xx integer;

set uid = event_parameter(慤ser�);

if exists (select * from dba.event_table

where ucase( uid) = ucase(username)) then

-- The user is already in the table.

update event_table set failed_login_attempts =

failed_login_attempts+1

where ucase(username) = ucase(uid);

else

-- Insert the user for the first time.

insert dba.event_table values (ucase(uid), 1);

end if;

select failed_login_attempts into xx from dba.event_table

where ucase(username)=ucase(uid);

-- It抯 not possible to stop the user from attempting to connect after 5 tries.

-- Instead, send a message to the server console to notify the database

-- administrator that a user has exceeding the allowable connect attempts.

if xx > 5 then

raiserror 17001 uid + � has had more than 5 failed login attempts.�;

end if;

end

The login procedure.

Create procedure dba.check_logins()

begin

declare xx integer;

declare uid char(128);

-- See if the connected user is in the event_table and proceed accordingly.

if exists (select * from dba.event_table

where ucase(username) = ucase(current user)) then

select failed_login_attempts into xx from dba.event_table

where ucase(username) = ucase(current user);

if ( xx >= 5 ) then

raiserror 17010 current user + � has been locked out by the

Database Administrator.�;

else

-- The user has connected.

-- Remove the user抯 row from dba.event_table and call

-- the default login procedure for the database.

set uid=current user;

delete DBA.event_table where ucase(username) = ucase(uid);

call sp_login_environment();

end if;

else

-- The user is not in dba.event_table, but has connected. Call the default

-- login procedure for the database.

call sp_login_environment();

end if;

end;

To enable all users to run the login stored procedure and set the login procedure option.

grant execute on dba.check_logins to PUBLIC;

set option PUBLIC.Login_Procedure = 慸ba.check_logins�;

To enable a user that has exceeded the number of allowed failed connection attempts to connect, a database administrator must delete the row for that user from dba.event_table.

functionality.

The rank functionality would need something like:

rank () over ( [query_partition_clause] order_by_clause )

same with dense_rank

dense_rank () over ( [query_partition_clause] order_by_clause )

This partition clause would really be similar to a group by clause.

queries where there is a large number of rows in the table.

queries where there are many rows in the table.

queries where there is at least one additional restrictive

predicate on a column from the same table.

drop table t2;

create table t2(c1 int, c2 int, c3 int);

create hg index t2c1c2hg on t2(c1, c2);

insert t2 values(1, 2, 3);

insert t2(c3) values(1);

delete from t2 where rowid(t2) = 2;

This is fixed in 12.6

create table t1 (c1 int, c2 int, c3 int)

create hg index hg1 on t1(c1)

create hg index hg2 on t1(c1,c2)

sp_iqcheckdb 'verify index hg1 index hg2'

The failure occurs because the two hg indexes overlap on a column.

Users with DBA privileges will be able to connect if either table is dropped. A message will be output to the server console to notify the dba or operator that non-DBA users will not be able to connect until the table(s) are put back.

The IQ_SYSTEM(USER)_LOGIN_INFO_TABLE cannot be found or accessed.

Non-DBA users will not be able to connect until the table can be accessed.

Put in for an option IQ that would allow for truncation of the iqmsg log such as "IQMSG_TRUNCATE"? The option would have 2 values, the default being 'NO'.

create table t1 (c1 int)

insert t1 values ('9')

will insert the row.

1. After executing the DDL statements to modify

procedures and functions at the write server,

execute the same statements on each query

server. Create a SQL script for this approach

to ensure consistent definitions.

2. After executing the DDL statement(s) to modify

procedures and functions at the write server,

synchronize each query server.

ALTER and DROP statements for procedures and functions can fail if the existing definition is in use by another connection to the server. The user should schedule maintenance of procedures and functions for periods when the existing definitions will not be used. This information was omitted from the Sybase IQ System Administration Guide.

has been improved for some queries where there is at least one

additional very restrictive predicate on a column from the same

table.

in MPX User's Guide is incorrect. It is necessary to add -x switch.

Issue was the result of improper handling of

expressions containing a varying host variable

plus constants or additional host variables,

but containing no columns or aggregates.

create table "t1""t1" ( c1 int )

will fail with a syntax error without the fix that this CR provides. This statement should create a table, t1"t1, with a single column, c1.

This enhancement will allow the load to continue when such an error occurs. Rows where data to be loaded into a Word Index is longer than the specified Word Index LIMIT will not be loaded into the table and the rows where data does not exceed the maximum permitted word size will be loaded. If a row is not loaded because a word exceeds the maximum permitted size, a warning will be written to the iqmsg file. WORD size violations can be optionally logged to the MESSAGE LOG file specified in the LOAD TABLE statement.

The existing behavior will be the default and the new behavior will be enabled by an option to the LOAD TABLE statement.

Here is the LOAD TABLE syntax showing this option:

LOAD [ INTO ] TABLE { owner.]table-name [ (column-name, � ) ]

FROM filename-string

[ load-option � ]

load-option:

[ CHECK CONSTRAINTS { ON | OFF } IGNORE ]

[ DELIMITED BY string ]

.

.

.

[ WORD SKIP number ]

� [ ONLY LOG logwhat [,匽 ]

WORD SKIP option : If the option is not specified, the LOAD TABLE will report an error and rollback on the first occurrence of a word that is longer than the specified limit. number specifies the number of times the 揥ords exceeding the maximum permitted word length not supported.� error will be ignored. 0 (zero) means there is no limit.

ONLY LOG option: The keyword WORD has been added to the list.

with an IQ Internal Error at location s_gUtil.cxx; 682. This has been fixed.

Customer needs to lock logins and finds the system stored procedures for managing logins are inconsistent and leave holes in security and user administration. They are suggesting the following modification:

"Grant Connect" and "Revoke Connect" do not properly reflect logins in the IQ_USER_LOGIN_INFO_TABLE. Information should be maintained there for all logins. "Grant Connect" and "Revoke Connect" should be wrappered in IQ to do this, or an option should be provided to disable direct access to those commands. The IQ User Administration feature is compromised by having several ways to get around it, intentionally or inadvertently.

We should be updating this table after issuing those commands for 2 reasons:

a) To be consistent with running the "sp_iqmodifyadmin enable" to update the IQ_USER_LOGIN_INFO_TABLE with logins added to the SYSUSERPERM. Why do we offer a way to update what should be the single point of storage for all logins anyway?

b) It is too easy to run "sp_iqmodifyadmin enable" and then issue a "Revoke Connect" thereby orphaning logins in the IQ_USER_LOGIN_INFO_TABLE.

c) To provide a single point of administering logins by using the "IQ User Administration.facility". Administrating/monitoring and controlling logins is currently too messy.

caused a failure on an internal consistency check.

java.io.IOException: Not enough space

This indicates an out of memory condition, not lack of disk space or dbspace. This was

omitted from Chapter 1, Sybase IQ Troubleshooting and Error Messages Guide.

query that accesses IQ store tables. The diagnostic information in

the message file shows the dbi_partialopen function at or near the

top of the stack trace.

In ESD1, the "Index Consultant" button is disabled in the dbisql

front-end tool that is distributed with Sybase IQ. If the SQL

Anywhere index consultant is started by other means, it will

run without a server failure.

Note that in general the SQL Anywhere Index Consultant will not

produce any useful information for Sybase IQ databases; the Sybase IQ

12.6 Index Advisor feature is the proper tool for automatically

generating index recommendations for Sybase IQ databases.

with the error "IOT/Abort trap".

The issue is resolved in 12.6 ESD1 by regenerating the Multiplex

Admin Scripts for the Multiplex Servers via Sybase Central.

ability to execute commands that require DBA privileges.

This example creates a policy that lets a non-DBA

user (user1) perform backup.

---------------------------------------------------------

create procedure "DBA".do_backup()

begin

backup database

crc on

attended off

block factor 4

full

to 'fileA' size 2000

to 'fileB' size 2000

to 'fileC' size 2000

;

end;

grant execute on "DBA".do_backup to user1;

Only indexes in base tables or global temporary tables with owner type USER can be altered. This information was omitted from Chapters 5 and 6 of the Sybase IQ System Administration Guide, from Chapter 6, Sybase IQ Reference Manual, and from Chapters 3 through 6, Sybase IQ Troubleshooting and Error Messages Guide.

changed" warning, then Open Client would fail to open the cursor. This problem has

now been fixed.

server failure. This issue is resolved and "comment" on

local temporary table is now prohibited with an informational

error "Cannot add a comment to a table declared as local temporary".

Workaround: Use the ODBC Data Source Administrator to remove DSNs created in 12.5

and add new DSNs.

Java in a database, there was a chance that a SQL Error could

result in server failure.

in this release", there was a chance server failure would occur.

the .iqmsg file (for example, via the unix

command cp /dev/null <iqmsg file>)

to clear out old information in the file. This let the

customer collect new information from the .iqmsg

file at regular known intervals to help

monitor the database engine.

The ability to truncate the .iqmsg file in this manner

was accidentally removed when the .iqmsg file wrapping functionality

was added to Sybase IQ.

New behavor:

> start_asiq @asiqdemo.cfg -f asiqdemo.db

Starting server demo_8888 on iqsunbld at port 8888 (02/08 09:23:53)

...

Database recovery in progress

Last checkpoint at Fri Feb 04 2005 15:54

Checkpoint log...

I. 02/08 09:24:03. Transaction log: asiqdemo.log...

I. 02/08 09:24:03. Forcing recovery without transaction log

I. 02/08 09:24:03. Rollback log...

I. 02/08 09:24:03. Checkpointing...

I. 02/08 09:24:03. Starting checkpoint of "asiqdemo" (asiqdemo.db) at Tue Feb 08 2005 09:24

I. 02/08 09:24:03. Finished checkpoint of "asiqdemo" (asiqdemo.db) at Tue Feb 08 2005 09:24

I. 02/08 09:24:03. Recovery complete

---IOX: Could not open script file

Wrapping iqmsg: Disabled.

Transaction log: asiqdemo.log...

Forcing recovery without transaction log

Rollback log...

Checkpointing...

Starting checkpoint of "asiqdemo" (asiqdemo.db) at Tue Feb 08 2005 09:24

Finished checkpoint of "asiqdemo" (asiqdemo.db) at Tue Feb 08 2005 09:24

Recovery complete

I. 02/08 09:24:04. Database server stopped at Tue Feb 08 2005 09:24

Database server stopped at Tue Feb 08 2005 09:24

Recovery complete

Make sure to remove the '-f' parameter before restarting the server.

in Sybase IQ 12.6. This statement prevents other concurrent

transactions from accessing or modifying a table. Syntax

is:

LOCK TABLE table-name [WITH HOLD] IN {SHARE | EXCLUSIVE}

MODE

Parameters are:

table-name - Must be a base table, not a view.

WITH HOLD - Holds the lock until the end of the connection.

Otherwise, releases the lock when current transaction is

rolled back or ends.

SHARE mode - Other transactions cannot modify the table

but have read access. Requires SELECT privileges.

EXCLUSIVE mode - Other transactions cannot execute queries,

updates, or any other transaction against the table. Requires

DBA authority or ownership of the table.

This statement blocks or delays other transactions that may

require access to the table. LOCK TABLE allows direct control

over concurrency at a table level, independent of isolation

level.

explicitly dropped or implicitly timed out and a

different user interrupting a command just scheduled

to run by the IQ resource gate could result in a

deadlock.

to the .iqmsg file. The new behavior is that each thread doing data conversion for a LOAD statement will post at most one data conversion warning message to the iqmsg file if

the database option "conversion_error" is set to "OFF".

In order to have all of the data conversion warning messages sent to the iqmsg file

an option must be set. This option is "DDL_OPTIONS2=8". Here is an example of how to

have all of the data conversion warning messages written to the iqmsg file:

set temporary option conversion_error='off';

set temporary option ddl_options2=8;

an insert to the bottom table of a join caused query server failure.

user at an unlikely but problematic point during the

first few seconds of execution.

is also a GROUP BY clause could result in either

an internal error from dfo.cxx:6235 or a Max_Cartesian_Result

error from dfo_Root.cxx:785.

during replay of TLV log on the query server. This happened

after upgrading the 12.5 multiplex environment to the 12.6 ESD1 release.

attempts to reference a view that needs to be recompiled

because of changes to the schema of a table used within

that view.

large tables, the thread allocation policies for predicate

execution have been improved. If you have lots of queries

that meet the criteria above, and especially if you have

such queries and are running on a system with multi-core

CPUs, then you should:

set PUBLIC.Enable_Thread_Allowance = 'ON'

Non-Fatal IQ Internal Error Detected at dfe_BaseColumn.cxx:886 on thread 34 (dfe_BaseColumn::GetMaxValue(void))

The view met the following criteria:

1) A UNION ALL view over ungrouped base tables

2) Each of the view's select list items mapped to a set

of identically typed base columns

3) One (or more) of the arms of the UNION referred to an

empty base table

4) The query block referring to that view performed

a GROUP BY or a join on a column from that view, such

that the optimizer requested the distinct count

of that column from the view

5) That view column had an unscaled integer

data type

6) All of the base columns mapped to by that view column

had HG indexes defined

7) The sum of the distinct counts for all the base

columns mapped to by that view column exceeded

~1.5 million

or SELECT DISTINCT on expressions could be rejected with an

"Internal error" that stated "Can not evaluate expression".

when trying to restart a query server without synchronizing

if the server had not cleanly shut down.

less efficient than they could be, causing queries to use more temp space and

execute more slowly than necessary.

The following cases were improved for queries where <col1> contains NULL values:

select ... where <col1> IS NOT NULL,

select ... where <predicate> AND <col1> IS NOT NULL,

select ... where <predicate> OR <col1> IS NOT NULL, and

select ... where <predicate> AND <col1> IS NULL

continuously grew when executing a multiplex usage scenario

that had a query server performing only queries and the write

server performing a repeated pattern of insert/load commit

into one or more tables, always modifying the tables in the exact

same order.

SET TEMPORARY OPTION identity_insert = ''; (This problem was fixed but undocumented in ESD 4.)

and concatenate slowed down when processing a large number of rows.

server fails when starting the IQ monitor

with the '-summary' option. For example:

iq utilities main into DBA.dummy start monitor '-interval 10 -summary'

the end of input in loads, inserts, updates, and queries. For

example, it treated "666 7" as "666".

There is a related IQ 12.6 CR 362288.

In 12.6.0 and prior versions, A, B, and C all had to be

constants, host variables, or columns.

In 12.6.1 either "B" or "C" or both "B" and "C" can now

be subqueries. "A", the left hand argument, must still be

a constant, a column, or a host variable in 12.6.1.

when inserting into a global or local temp table

could be reported performing an UPDATE operation on a HG index.

names depending on how the procedure is called form isql. When called as

"call sp_iqdbspace()" it returns one set of column names, when called as

"select * from sp_iqdbspace()" it returns a different set of names.

server failure reported at s_blockmap.cxx; 974.

The following are the results for bigint, integer and numeric values:

'00000001' '00000001' '00000001'

'ffffffffffffffff' 'ffffffff' 'ffffffffffffffff'

'00000002' '00000002' '00000002'

'fffffffffffffffe' 'fffffffe' 'fffffffffffffffe'

'7fffffff' '7fffffff' '7fffffff'

'ffffffff80000001' '80000001' 'ffffffff80000001'

'ffffffff80000000' '80000000' 'ffffffff80000000'

'80000000' '00008000' '80000000'

'7fffffffffffffff' '00007fff' '7fffffffffffffff'

'8000000000000001' 'ffff8000' '8000000000000001'

'8000000000000000' 'ffff7fff' '8000000000000000'

This problem has been fixed. The following shows typical results now returned for bigint, integer and numeric values:

'0000000000000001' '00000001' '0000000000000001'

'ffffffffffffffff' 'ffffffff' 'ffffffffffffffff'

'0000000000000002' '00000002' '0000000000000002'

'fffffffffffffffe' 'fffffffe' 'fffffffffffffffe'

As well, with 9.0.2, passing a negative value to inttohex() would have resulted result in a "value out of range" error. This to has been corrected

unresponsive when deleting an entry from an invalid index instead of reporting

Assertion 200901.

may cause the query server to stop with an emergency shutdown message.

may cause the query server to stop with an emergency shutdown message.

of tables in the database that is being opened on startup.

1. The "Show multiple result sets" option was turned on, and

2. The "Maximum number of rows to display" option was set to "All rows", and

3. The result set contained a lot of data.

can occasionally fail with the engine reporting an IQ Internal

Error - s_ohcidx.cxx; 1379.

Workaround: using -iqmpx_sn switch.

and using an IN predicate for the pushdown caused a server failure.

when the write server is being changed to single-node mode while

another write server connection is running active transactions and generates tlv entries.

for a tinyint column if either 'Minimize_Storage' = 'ON' or a

column unique count is specified in the rebuild index.

create variable @chkey decimal(2);

select @chkey = 1;

select @chkey FROM iq_dummy;

a delete operation could cause the engine to fail with a SEGV or Signal 8 error

depending on the structure of the FP indexes.

due to locking of SYSIQFILE system table.

join conditions or uncorrelated IN subquery

conditions slowed down between ESD3 and ESD4.

might occur before a very long array or string was allocated or while pieces of that array/string were being allocated. In this situation, if the Garbage Collection ran multiple times, the IQ server could fail. This typically occurred if a user ran a java

program inside the IQ server or used CIS to connect to remote

servers using jdbc connections and the program was memory intensive.

portion of a server shutdown could report the error "Assertion failed: 101416

Heap left locked at database close" in the server log.

greater than the column being queried could cause server failure.

A LIKE query on a column defined as char(5)

with a string of six or more characters is an example of such a query.

while replaying a BEGIN PARALLEL IQ clause that

had previously successfully executed on the write server.

could report DBCC Unknown Exceptions.

join key column with more than 700 million distinct values

could, under unusual circumstances, cause a server failure.

cursor operation when no data had been previously

fetched.

of a updatable cursor operation.

unexpected results.

implemented in 12.6 GA, but it was omitted from documentation. It is now documented in the Sybase IQ Release Bulletin.

trying to determine if CASE expressions in the ORDER BY

clause and the SELECT list match when there is an

IN condition.

was reported during a large delete.

Invalid database server command line

replaying a ALTER TABLE RENAME command propagated

from the write server to a query server.

1) Connection 1 ran a stored procedure.

2) Connection 2 dropped a table used by the stored procedure in

Connection 1 and recreated it as a view.

3) Connection 1 tried to run the stored procedure again.

+ the path to the 3rd-party ODBC driver library is included in $LD_LIBRARY_PATH_64;

+ the IQ startup script is modified so that the path to the 3rd-party ODBC driver library is included in $LD_LIBRARY_PATH_64.

is started while it is executing.

performed on char, varchar, or varbinary

columns with a defined width greater than 255

bytes or clob/blob columns whoses data widths are less

than 32768 bytes may cause FP index inconsistencies.

The inconsistencies may cause non-fatal s_blockmap

errors.

The issue is extremely data pattern dependent and is

much more likely to occur when updating more than 10,000

rows in a single update command.

The issue does NOT occur if any of the following are true:

- All rows in the column are updated in a single

update command.

- All rows being updated were originally all

NULL or zero-length strings

- All rows being updated are set to NULL or

zero-length strings.

- It is a single table update (no joins involved)

whose WHERE clause specifies single-column

predicates against constraints.

Once you have applied 12.6 ESD6.2, run the following

command to check the FP indexes:

sp_iqcheckdb 'check database indextype FP'

If no errors are reported, then your database was

unaffected by the issue and no further action is

required.

If inconsistent FP indexes are reported, run the

following command for each inconsistent index:

sp_iqcheckdb 'check index index_name'

A list of affected rowids will be printed in the

.iqmsg file preceded by the following line:

DBCC owner.table.indexname: # inconsistent ROWIDS

(where # is the number of rowids)

To resolve the inconsistencies, the following steps

must be repeated for each inconsistent FP index

reported. The sp_iqindex procedure can be used to

determine the column name that corresponds to the

FP indexname.

1) Add a new column to the table with the same

definition as the original column with the

inconsistent FP index. For example:

ALTER TABLE table_name ADD new_colname type

(width) NULL

2) Update the new column with the data from the

original column excluding all rows reported

reported in the .iqmsg file under 'inconsistent ROWIDS'.

For example:

UPDATE table_name SET new_colname =

original_colname WHERE ROWID(table_name)

NOT IN (list of inconsistent rowids)

3) Match each rowid reported in the 'inconsistent ROWIDS'

list to the table's primary key or, if no primary

key exists, use data values from other columns

that in combination uniquely identify the row

Use this information to determine the correct

column value from the original data source used

to populate the column. Update the new column

based on this information. For example:

UPDATE table_name SET new_colname = 'value' WHERE

ROWID(table_name) = rowid value

4) After step 3 is complete for each inconsistent

rowid reported, drop the original column and

rename the new column to the original name.

For example:

ALTER TABLE table_name DROP original_colname;

ALTER TABLE table_name RENAME new_colname

TO original_colname

If the original column was defined as NOT NULL,

alter the column to add the NOT NULL

specification. For example:

ALTER TABLE table_name MODIFY

original_colname NOT NULL

5) If the original column had WD or CMP indexes,

recreate them.

6) After steps 1 through 5 are completed for

each inconsistent FP index reported,

run sp_iqcheckdb 'check database indextype FP' a final

time to confirm that no errors are reported.

on SA versus IQ tables. Both should return the base data type. (Duplicate of CR 442804.)

or incremental insert to a char, varchar, or

varbinary column wider than

255 bytes, the server may fail with a

"Destroying Locked Logical Buffer" error.

procedure and the procedure contained statements of the form "SELECT ... INTO #temp ...".

UID=DBA

PWD=SQL.

This backup database command fails: backup database to 'c://temp//test.bk';

But works in ASE.

returned a SEGV error.

a Fatal Error could sometimes be returned when SA tmp (tmp_00000000)

filled the disk, depending on the cause of the temp space increase.

value of 3, multiplex synchronize might cause a server Signal 11

failure.

same stored procedure for the first time after the database start and the procedure

was used in a select statement.

the database option "CONVERSION_ERROR" is set to "OFF".

In order to have all of the data conversion warning messages sent to the .iqmsg file

an option must be set. This option is "DDL_OPTIONS2=8". To write all of the data conversion warning messages to the .iqmsg file, use these statements: SET TEMPORARY OPTION CONVERSION_ERROR='OFF'; SET TEMPORARY OPTION DDL_OPTIONS2=8;

FROM js1 LEFT OUTER JOIN( SELECT DISTINCT(col_1) AS col_1, 'R' AS col_3 FROM js2 ) AS t_js3 ON js1.col_1 = t_js3.col_1

GROUP BY t_js3.col_3.

even if the cursor was not declared for update.

for the derived table. If the column names of the derived table

happened to be the variable names in the subsequent arms, the query could not complete and the server failed.

might be incorrect if an index that maintains cardinality (e.g., LF or HG)

was created on the writer after the query node has been

created or synchronized.

WHERE clause with the ROWID function could cause server failure.

when the date_order option is set to 'DMY' in the 12.5 database.

with ROWID could fail with an error from vpp_IsNull.

value expression, and where that expression contained a column

which was used as a join key at a lower level join, then that

query could under some circumstances be incorrectly rejected

with a non-fatal internal error stating that the input to a join

operator was "not ordered as required".

upon restore from backup the uncommitted changes may not be rolled back for all connections.

This could produce catalog inconsistencies and subsequent server failures if a command

which creates or alters a table index was running during the backup.

with advice to rebuild a column with MINIMIZE_STORAGE='ON' even

when MINIMIZE_STORAGE was ON and a 1 or 2 byte lookup FP index existed.

non-DBA connection could roll back despite the CHAINED option being set 'OFF'.

Please access the following URL, when you use DBD::ASAny with IQ.

http://www.sybase.com/detail?id=1009063

But DBD::ASAny must install ASA. This driver cannot be used only by IQ.

in a join index, then subsequent operations involving tables in the

join index could report an error from db_joinindex_id.cxx.

Release 12.7 or later, sp_iqrelocate might report an -1006002

"IQ internal error" on any table that contains a column with

NULL values, if that table has not been modified since the

upgrade.

a view or derived table in the FROM clause that used the

same table alias as the delete table, may delete rows from

the view or derived table with the duplicate alias.

In order for the issue to occur, all of the following

conditions must be true: 1)The DELETE statement has a FROM clause with multiple entries and one of those entries is a view or derived table; 2) An identical alias is used for both the delete table and a table in the view or derived table; 3) The delete table precedes the view or derived table in the FROM clause; 4) The view or derived table does not contain a DISTINCT clause, GROUP BY clause or any aggregates in its select list; and 5) The DELETE statement determines there are rows to be deleted based on the statement criteria.

that contains NULL values, an error from dfe_MinMaxAggregate.cxx could be reported.

return fewer rows than specified.

range and the PBV algorithm was selected.

To repair this,start the server with the '-iqnotemp' switch and drop the affected temporary dbspace. Then shutdown and restart the server normally.

console.

and a date comparison in conjunction with an OR operator could cause server failure.

set to a value larger than 30 and the server accumulates at least 30 distinct advice messages.

- a stored procedure references a user-defined function in the search-condition of an IF statement

- the function calls a procedure with OUTPUT or INOUT parameters

- the function then calls another procedure

could report a non-fatal error "cannot evaluate expression"

depending on the cardinality of the data.

where the derived table involves both local and remote tables

may cause server failure.

then server failure occurred instead of reporting the

error that ROWID over a view is not supported.

the second time the application was executed, a "cursor not open" error was reported.

is the product of the -iqmt value multiplied by the -iqtss value.

Using dbisqlc for IQ 12.5 ad using IQ 12.5 same thing works without any problem. On IQ 12.6 and 12.7 it fails. So it seems to be problem intoduced in IQ 12.6 and up.

Problem is not observed when connected to the IQ server from same host machine on which IQ is residing.

conditions, where one of the join keys is equated to

a constant value expression in another condition and

where the optimizer selected either the SMJ or the

SMPDJ algorithm for the join, then that query could

under some circumstances be rejected with a non-fatal

internal error stating that the input to a join

operator was "not ordered as required".

where each arm is grouped may return extra rows.

Can IQ be improved to see if there may be a way to better handle this condition so that IQ is not killed abruptly and cause recovery issues / problems. Could IQ issue a warning, or would there be a way to tell the OS not to select it, or even if IQ tries to dynamically reduces its requirements to alleviate the problem. Obviously this would be dependant on what the OS does, i.e. its not clear yet if the OS issues the SIGDANGER(33) then immediately start killing processes, and if so what signals does it send.

We are already aware that the underlying condition can be avoided by configuring more memory / paging space ... etc. etc. However in reality its not always possible or desirable for a customer site to provide excessively large amounts of these resources, so some degree of tuning is done based on 'normal' requirements, hence from time to time customers can and do hit this issue. Can we make it a little less painfull for them?

used a 32-bit file path when entering the service.

while executing operations that update HG indexes or LOB columns after

adding a dbspace that was assigned block numbers larger than 4 billion.

might fail with the error "This application has failed to start because

the application configuration is incorrect. Reinstalling the application

may fix this problem."

the -iqro 1 startup option failed.

a query which involved a view which contained a UNION ALL of

2 SELECTS in the FROM clause and a WHERE clause with a

BETWEEN condition on 2 dates and an OR condition.

ran the same procedure concurrently and the procedure contained a

control statement (for example, IF) with a condition that used a subselect.

not have local stores anymore, eliminating this kind of name collisions causing emergency

shutdown of the query server.

reported an "Incompatible data types in UNION" error.

first access of the global temporary table.

join took a long time to be optimized.

containing null values that was created in 12.6 or earlier and that column

had not yet been modified in 12.7.

creation of the proxy table failed.

select col1, col2 from table1 where

col1>=ALL( select col1 from table2 where col1=1 )

order by col1, col2;

might occur, if dml_options5=1073741824 was set.

failure could occur.

involving a join and a NOT IN subquery where the subquery could

infer the predicate from a join condition could return unexpected

results, since the "NOT" condition was dropped. In order to infer

the predicate from the join condition, the data types of the join column

and subquery column must be identical.

instead of the expected NULL value under the following conditions:

1) There were no GROUP BY items associated with the MIN/MAX.

2) The argument of the MIN/MAX could be simplified into a constant.

(For example: SELECT MAX(a) FROM t WHERE a=6)

3) No result rows satisfied the query predicate.

could occur when connection pooling is enabled and many users are concurently connected.

block usages increasing, until the cursor was closed and the connection

was disconnected. The workaround is not to use the cursor WITH HOLD

option.

or VARBINARY UNION ALL view column as a distinct or grouping column

could return a non-fatal internal error.

Sybase IQ 12.7 ESD #3 clients and servers have been updated to use Certicom v5 libraries

for Transport Layer Security (TLS) encryption, HTTPS encryption, and aes_fips database

encryption on the following platforms: AIX64, Linux AMD64, Sun64, and Windows 32-bit.

This update changes product behavior in a single instance. To connect a version 12.7

ESD #3 Network Client to a 12.6 Sybase IQ server encrypted, you must use the RSA

protocol. Attempts to connect encrypted to unpatched or 12.6servers using the Elliptic

Curve Cryptography (ECC) protocol return error code -829: TLS handshake failed.

when the remote table name contained an underscore would fail

with the error "The remote table <table name> could not be found."

(Duplicate CR 454785)

option was set to the non-default value of 'ON' and a client-side cursor

was closed and re-opened. The cursor operations could be called explicitly

from a client library or as the result of a CIS decomposed query

or OUTPUT TO statement.

of an ASE IMAGE column or IQ LONG BINARY column to an IQ

BINARY or VARBINARY column.

cached stored procedure.

Successful connections by host name might fail when the IP Address is used.

in a GROUP BY clause could cause the server to be unresponsive.

create table;

insert table;

commit;

open cursor for select from table;

commit;

insert table;

fetch cursor;

commit;

insert table;

commit;

close cursor;

See the Release Bulletin for a description of related behavior changes.

control the degree of parallelism used for UNION query processing.

rows are updated or deleted. ODBC SQLROWCOUNT will be -1 instead of 0 for these two

cases.

CREATE foo (c1 INT);

through a database stop and restart. If the IQGOVERN_PRIORITY

and IQGOVERN_MAX_PRIORITY options were set to 1 for the PUBLIC

group, the warning "Option IQgovern_priority has an invalid

setting: 1." was reported on database restart.

backup that moves the catalog store.

temporary tables could become unresponsive.

"ASA Error-300" and cause a stack trace.

will probably cause the IQ temporary store to fill up.`

the correct percent of the dbspace used by the index.

the message "Key too large" or insert ... location ...could fail with

"Page size selected is too small for this datatype".

block (this is not allowed) would not close the cursors that

were created for the SELECT. This would cause the server to fail

at st_txncb.cxx line 2822.

complexity exceeds server limits."

INOUT). Instead, it returns 0 for all the modes. This has been fixed.

SELECT col1 from remote_procedure_1() where col2 = (select c1 from remote_procedure 2())

freshly created IQ database using a raw partition on Windows.

when a period is present, instead of a comma, between two items in a SELECT list in

which the second item is an aggregation function.

INSERT...LOCATION, SELECT...INTO, LOAD TABLE, UPDATE, TRUNCATE, DELETE, SYNCHRONIZE JOIN

INDEX, and updatable cursor operations.

Now IQ will look into both /dev and /dev/raw to find the rawctl file.

... then if you try and connect from SC to the IQ server using an ODBC data source file (.dsn file), the connection works but you are not able to open the multiplex folder.

The length limit of 254 bytes is now enforced and a new error message "Function not

supported for the string > 254" is reported.

This issue relates to CR#480255.

select id, fname into #ltt from customer where id = -99;

commit;

create index hg0 on #ltt(id);

savepoint S4;

create index hg2 on #ltt(fname);

drop table #ltt;

- a cursor is opened on a procedure call that returns a result set

- a parameter to the procedure is a user-defined function call

- the UDF issues a COMMIT

is not a SELECT), the server could fail. This issue could also occur if the SELECT is

used as a subselect in a DML statement. This issue is resolved.

This appears to be a dupe of cr 466135.

the error "Data exception. Data conversion is not possible."

User cannot uncheck checkkbox "No default or computed value" on "Column property -> Value panel" , after checking it.

with a HAVING clause reported Internal error at dfo.cxx

(Cannot evaluate expression).

data cannot be loaded successfully if the data contains quotes. For example, if a column

contains a'b'c, the extract file contains 'a'b'c', which cannot be loaded by the

LOAD statement with QUOTES ON.

columns could cause server failure in dfo::DFEMightBeNullHere().

for Windows did not match the IQ build version. This issue has been resolved.

� The whole query text is now logged in the request log as compared to current limit of 32k.

� If the output of logging is directed to server log/console then only 32k size of query text will be logged.

� Stored procedure sa_get_request_times will truncate the request log record to 64k size.

Workaround: Ignore the references to ESD#1 it should read ESD #2.

Workaround: The user can delete the installation from the install directory manually.

1. inserted N (an integer or an exact numeric type with scale 0) to the DEFAULT AUTOINCREMENT/IDENTITY column, and

2. inserted N-1 to the DEFAULT AUTOINCREMENT/IDENTITY column, and

3. tries to insert a value to another column of the table.

verification tests. This has been fixed.

i.e. who is associated with the failed event being audited

with this fix, the query plan displays a more meaningful message "Statement text unavailable".

select * into #temp_t1 from sp_iqstatus();

select * from #temp_t1;

This is incorrect and rows will be returned in 12.7 esd5.

file is available with proper permissions in lib directory.

user1:

create table t1;

lock table t1 in shared mode;

user2: (who does not have select permission to user1.t1)

lock table user1.t1 in exclusive mode;

The lock table attempt by user2 used to report error "Permission denied: you do not have permission to alter t1.". In 12.7 ESD #4 and later, the lock attempt now returns error "User 'user1' has the row in 't1' locked."

create table test_table (c1 int);

go

insert into test_table values (1);

go

alter table test_table add c2 bit not null default 0;

go

select * from test_table;

go

The error is:

CT-LIBRARY error:

ct_fetch(): protocol specific layer: external error: There is a tds protocol error. An illegal token was received.

===Fix by Engineering=== ( Imran )

IQ server crashed on "Group by" while costing was calculated as security_code_cookie check was inserted by MS VC for windows 64 bit platform. Care has been taken security cookie is not set any more for this function.

error. This has been fixed.

Now identity values are generating correctly.

list of a select INTO <permanent_table> statement, the statement fails with the

"Syntax error near '0' on line 1".

than or equal to number of CPUs. When MAX_QUERY_PARALLELISM is set, the number of parallel

arms in a UNION GROUP BY can be greater than the number of CPUs.

select count(*) from ... where (((... is null) or (... in ...)) and ...)

may cause the IQ server to crash. This is due to an attempt to write to a

read-only bock map. In this case, a copy of the block map needs to be made

and used. This has been fixed.

This problem occurred while estimating cost of a correlated sub-query. The fix resolves cost estimation issue and query now runs without error.

What's needed is the x64 version of Psapi.lib to build findproc.cxx. This has been fixed.

CUSTOMER NAME: RBC DOMINION SECURITIES

ASIQ Version: IQ 12.7 esd #3

PLATFORM and OPERATING SYSTEM: Solaris

BASIC INFORMATION:

Customer will like to have some means that will show table names and blocks usage in TEMP space store (like ASE).

As a part of fix we raised an exception string and report the value returned by the call to pthread_mutex_destroy , in this way, we are ignoring the EBUSY return .

64bit installation.

data types only. They are not defined for bit type. Because

of this, the input operand of the bit negation operation is

casted to integer first. After applying the bit negation, we

try to cast the resulting integer value to bit again, which

does not work.

Default.ini file. The keyword is 'timeout'

and it belongs in the Network area of the

.ini file. The valid values for timeout

are specified in seconds in a range from

1 second to 2,147,483 seconds.

An extract of the Default.ini file contents

shows the appropriate syntax. For more

details on default.ini contents, refer to

the User's Guide.

==Begin snippet of default.ini======

# The timeout setting must be in the

# Network section. The beginning of the

# Network section is defined by the line below.

[Network]

timeout=3600

=End snippet of default.ini===========

Please note this timeout affects only

the ETL software acting as an ftp client.

This timeout value will not affect any

timeout setting that are defined on the

ftp server or enforced by the network at

the TCP layer.

High FS cache on Red Hat 4 and Red Hat 5 impacts concurrent transactions on IQ 12.7.

FS-Cache is a kernel facility by which a network filesystem or other service can cache

data to improve performance. Because the current transaction version in IQ 12.7 is

released only after the FS cache is totally flushed to disk, a high FS cache can

impact concurrent transactions.

Reducing the vm.dirty_background_ratio and vm.dirty_ratio parameters increases the

number of flushes, but decreases the duration. Shorter flush times release the current

IQ version more quickly and reduce performance problems.

Note

-- Red Hat 4 has no configuration parameter to limit FS cache size. In Red Hat 4,

FS cache depends on machine memory.

-- With Red Hat 5, this should be worked around by enabling direct I/O on the OS

level to disable FS caching (not available in RH4).

select * from (select top 2 cust_id = rowid(t) from sales_order)t;

a Sybase IQ Abort error.

As an example (in pseudo code):

drop table xxx;

create table xxx( c1 int );

open curs1 { select * from xxx };

insert into xxx values(1);

open curs2 { select * from xxx };

insert into xxx values(2);

open curs3 { select * from xxx };

insert into xxx values(3);

open curs4 { select * from xxx };

commit;

print 'cursor1 ======'

loop {

fetch curs1 relative 1 into c1

if SQLCODE = SQLE_NOTFOUND then { break loop } endif

print 'c1 = ', c1

} endloop

print 'cursor2 ======'

loop {

fetch curs2 relative 1 into c1

if SQLCODE = SQLE_NOTFOUND then { break loop } endif

print 'c1 = ', c1

} endloop

print 'cursor3 ======'

loop {

fetch curs3 relative 1 into c1

if SQLCODE = SQLE_NOTFOUND then { break loop } endif

print 'c1 = ', c1

} endloop

print 'cursor4 ======'

loop {

fetch curs4 relative 1 into c1

if SQLCODE = SQLE_NOTFOUND then { break loop } endif

print 'c1 = ', c1

} endloop

When running the script, the following results are produced

cursor1 ======

cursor2 ======

c1 = 1

cursor3 ======

c1 = 1

cursor4 ======

c1 = 1

c1 = 2

c1 = 3

The results for cursors 1, 2, and 4 are correct. The results for cursor3

are incorrect. Cursor3 needs to see rows with values 1 and 2; not just 1.

The dbspace_online column was added to the output of the sp_iqdbspaceinfo stored procedure. A value of Y means that the dbspace is online and that the procedure returned data for that row. A value of N means that the dbspace is offline, meaning that the procedure was unable to return data for that row, and reports 0B (0 bytes) for those data values.

The dbspace_id column was added to the output of the sp_iqdbspaceobjectinfo stored procedure.

If you upgraded the server from a version earlier than 15.1 esd 1, you must run ALTER DATABASE UPGRADE on existing databases. If you do not upgrade the database, running the sp_iqdbspaceinfo stored procedure results in a 揚rocedure no longer valid� error.

12.7 Usage: stop_asiq -version [12|15|all]

15.1 Usage: stop_iq -version [12|15|all]

multiple cursors are being opened and destroyed and information on cursors is being

collected by sp_iqcontext.

rapidly with Txn-Cmt-PostCmt messages.

LINUX requires that the size of the transfer for read and write must be an integral multiple of the device's blocksize. For a disk, this is its sectorsize, which is normally 512 bytes. The other valid sectorsize values are 1024 and 2048. A blocksize of 4096 will not work.

You may use any of these values: 512, 1024 or 2048. Then, provided raw partitions are correctly setup, IQ backup and restore are successful.

and large nested CASE statements was blank. Now the

output of the Join node contains the required entry.

The ETL Engine Monitor memory foot print grows over a period of time.

Resolution:

This was resolved by modifying our garbage collection, clearing unused objects and removing the saving of some unneeded diagnostic information in memory.

Sybase Central and DBISQL incorrectly handle object names that begin with a space character.

If a table (or view) is named "<space>customer", then Sybase Central drops the leading space when

communicating with the server. For example, trying to drop the table with SC results in

"drop table customer" being sent to the server rather than "drop table "<space>customer"".

Embedded spaces in an object's name are handled ok.

Tspaces are also handled ok. Table and view names are database column data

and trailing blanks in character data are ignored for comparison purposes.

For example,

create table "ab<space>" ( c1 iint );

create table ab ( c1 int ); <--------------will get table already exists

drop table ab; <-----------------------------will drop "ab<space>"

An exception is:

create table "ab<space> ( c1 int );

select * from ab; <-----------------will give an abiguous table name error when we should do the select.

<linked_server_name>..dba.myproc", Microsoft SQL Server

2005 passes a statement of the form {?=call "dba"."myproc"

} to the SQL Anywhere OLE DB provider. It passes in a

single integer parameter for binding with a status of

DB_E_UNAVAILABLE. As such, the EXEC statement failed

with an error.

Resolution: ??????????????????????????????????????????????????????

chapter 10: Transactions and Versioning

How transaction information aids recovery

[Optionally after incremental backup.] > Is a transaction log truncated by incremental backup truly?

"HY000" is set to ODBC 2 or 3 State of these errors when ODBC or iAnyWhere JDBC is used.

However, description does not exist in a manual about "HY000."

create table t1 ( c1 date, c2 time );

insert t1 values ( '12/12/12', '05:05:05' );

select c1.foo(), max( c2 ), c2 from t1;

An error is now reported. "Invalid type for field reference."

on Linux.

created with CREATE PROCEDURE ... AT ... and calls a remote stored procedure residing on

another IQ server (e.g., IQ 12.7 ESD #2) may give an error like the following:

Msg 13732, Level 16, State 0:

ASA Error -660: Server 'sunspot': ASA Error -131: Syntax error near '.' on line 1

This has been fixed.

select top 10 act_rst_id into #aa from f_acct_rst order by prd_id

need a possible Feature Request to have ORDER BY optimization for TOP/FIRST put in

Here is the scenario:

create table mytab(att int) in SYSTEM;

insert into mytab location 'SYB125.base1' {select * from mytab};

Could not execute statement

Feature 'location' not implemented

SQLCODE=-134

Ct spent lots of time to find out that pb was due to the fact that the table was in the SYSTEM dbspace.

Could we get a message like

Feature 'location' not implemented in SYSTEM dbspace

instead?

1. Change the double quotes to single quotes

2. Include "set quoted_identifier off" before the SQL statement to be executed:

FORWARD TO <remote server name> {

set quoted_identifier off

<SQL statement>

}

3. Set the quoted_identifier to off using a separate FORWARD TO statement:

FORWARD TO <remote server name> { set quoted_identifier off }

ETL takes too long to set the query statement via ODBC on a huge table.

Resolution:

A change was done to delay the execution of the sql statement to when the project is run to avoid server load during design.

For example:

select MarketDateTime,LastPrice,count(*)

over (partition by NumericalInstrCode, BucketId order by LastPrice rows between

unbounded preceding and current row) PriceRank from test

where MarketDateTime between '2008-01-01' and '2008-01-02'

and NumericalInstrCode=1;

and the emergency buffer pool is involved. This is a duplicate of CR 273582 and has been

fixed.

When using Insert Location, some table names are truncated at 30 chars when browsing a database.

Resolution:

The code that gathers the table names was changed to use a SQL statement rather than sp_tables when the DBMS is Sybase IQ Version 12.n

Non-Fatal IQ Internal Error Detected at dfo_cOrderBy.cxx:217:

create table T1 (C1 varchar(10));

insert into T1 values ('Tabla 1');

create table T2 (C1 varchar(10));

insert into T2 values ('Tabla 2');

commit;

select top 10 * from (

select 1 as M from T1 union all

select 2 as M from T2

) as R order by M asc;

ETL throws an exception when retrieving data from MSSQL when "AUTO_COMMIT=0" is set.

Resolution:

The transaction handlers for ODBC were modified to recognize when the Auto commit setting is turned off.

In IQ itself (i.e. select * from SYSCAPABILITY) the column name for SYSCAPABILITY is srvid. However the IQ documentation manuals refer to the column name as svrid

ETL fails to properly read some dates when accessing databases using the ct-lib/Open client interface.

Resolution:

Changes were made to ETL's database interface class to properly handle column sizes.

select * into t from sp_iqstatus();

select graphical_plan( 'select * into t from sp_iqstatus()',1,'ASENSITIVE','READ-WRITE');

This problem is fixed and a proper error is now given:

-894 Plan cannot be generated for this type of statement.

When configuring DB Data Provider Full Load component using ASE15.0.2 ESD5 with ODBC, ETL will only see dbo in Schema field, other schemas are not present. Manually typing in the schema manually into the schema field with not work either.

Workaround:

There are 2 workarounds for this issue. Use ASE native connectivity instead of ODBC or use the ASE15.0.2 GA version. This will be fixed in the ETL product in future releases.

It is missing from Atomic GA.

The ETL UI hangs when opening a file that contains data for 10,000 columns.

Workaround:

When the OPEN dialogue opens, cancel it, click on the auto-update button in the component, then return to the OPEN dialogue, and open the file.

Wide tables (tables in Sybase IQ with 100s or 1000s of columns) use a lot of memory. To tune your target IQ system to handle wide tables, read the Sybase IQ Performance and Tuning Guide, especially chapter 5 Managing System Resources.

The following observations and recommendations are based on using Sybase ETL to migrate wide tables from one Sybase IQ instance to a target Sybase IQ instance with tables configured with varying numbers of columns and rows.

Observations and Recommendations

Repository:

- It is recommended to use a SQLAnywhere (SA) repository when using Wide Tables rather than an MS Access or a SQLite repository.

- It is recommended to allocate a 1 gigabyte database for the SA repository with all the other values remaining as the system defaults.

InsertLocation:

- Insert Location is supported using a 3000 column table.

Migration Wizard:

- Migration Wizard is supported using a 3500 column table with both OCS and ODBC interfaces.

Load Table:

- Load Table is supported using a 3000 column table with the OCS interface.

- Load Table is supported using a 10,000 column table with the ODBC interface, although performance suffers as larger numbers of rows are moved.

User Interface Experience

General Limitations:

When working with wide tables, the UI can take a long time to respond to user requests. In general, it is a good idea not to preview any data with this many columns. This could cause the UI to hang for long periods of time.

Insert Location Component:

- There are currently no known user interface specific limitations when configuring the Insert Location Component

Load Table Component:

- The Text Source dialog will truncate the display of data from files containing more than 50,000 bytes of data.

- If the data file contains more than 5800 columns, an error will occur and the output port preview data will not be available. This error will not prevent configuration of the component - the UI will continue to function normally without output port preview data. It is also possible that the UI will hang when previewing tables as wide as 10,000 columns.

Workaround:

- To avoid this, cancel out of the text file selection dialog, click the Auto-update button (to turn it off), and select your text source file.

Migration Wizard:

- Occasionally, the Migration Wizard will fail to generate all of the migration projects. This is usually due to memory limitations or using a non-SA repository.

Query Designer:

- Adding thousands of columns to a query at once can cause the UI to hang.

certain queries contained a reference to a null column.

involves a local temporary table, the server may fail.

UDF, and the UDF call passed in column references as

parameters, and the server ran the query in partial passthru

mode, then there was a chance of server failure.

not include an entry for one of the other query servers, resulting in the

"IQ_MPX_VERSIONLIST is missing row" error when sp_iqmpxvalidate is run after the writer

server replacement steps are completed. This has been fixed. In the 12.7 System Administration Guide, step 2 "Replacing a write server that has no SYSTEM dbspace and/or

transaction log file" on page 668 should read:

2 Choose the query server with the largest value in "DBA".IQ_MPX_STATUS.catalog_version

(or any one of those query servers, if the largest value exists in more than one query server) to be the new write server.

HP-UX for example), the server failed in

extremely rare conditions, when using a connection number to

get connection information. Examples of getting this type

of connection information include getting a connection

property for another user, or calling sa_conn_info.

and if the remote query contained a GROUP BY ... HAVING

clause, then using the view in a query caused a server

failure.

parameter was passed to dbisql and the -jconnect switch was

used, the client failed.

stack overflow' ..." error. The workaround is to use a larger catalog page size when

creating a new database, and is to split long statements into shorter ones for existing

databases.

A query server may not modify objects in the main store.

with following error codes:

SQLSTATE: QXA19

SQLCODE: -1010019

Sybase Error: 21113

The error message has been changed to following:

Cannot modify object <object name> on dbspace <dbspace name>. A query server may not modify objects in the main IQ store.

with following error codes:

SQLSTATE: QXA31

SQLCODE: -1010031

Sybase Error: 21131

This problem has been fixed.

Example. ./setupHP-UXi64 -is:jvmtimer 30

If the problem still persists try increasing the value from 30 to something larger.

Since the store procedure definition has changed, user will need to run ALTER DATABASE UPGRADE.

CAST larger than BIGINT.

by inversion for a query with a range predicate.

to a table with default values.

Here is a sample to illustrate the failure.

create table t1( c1 int, c2 date default current date,

c3 time default current time );

load table t1( c1 ascii(4), filler(1) )

from 'data.inp'

format binary

quotes off escapes off;

select count(*) from t1;

NOTE: datafile 'data.inp' must contain the following

1234

1234

The file gets created correctly but the TLV log has an incorrect entry. On replay the replay fails silently and the file is never opened

With this release, in this scenario, the restore operation may either succeed or give following error:

Cannot execute 'restore <dbspace or dbfile name>' since previous alter command is not yet propagated through the multiplex. Try again.

SQLSTATE: QSA66

SQLCODE: -1009066

Sybase Error: 20087

with a user interrupt, the reader became unresponsive due

to a deadlock.

The problem has been fixed.

the correct time. For example:

create table t1 ( c1 time );

insert t1 values ( current time );

select dateformat( c1, 'hh:mm:ss' );

Will output 09:mm:33

B an IQ object, caused an IQ server failure with a signal 11.

server option set after restoring the database. This

database was a Multiplex Reader database being started

using the "-iqmpx_ov 1 -iqmpx_sn 1" options as well.

Regarding follow-up, we are currently taking stock of our findings and roadmap options.

Dom Ruffatto is my product management counterpart, and will be the one to contact you regarding any follow-on discussions. I CC抏d Dom on this email, by way of introduction.

Dom � Peter Do is the Sybase account rep for HP.

Regarding testing results.. here is a very quick summary of my findings:

- Select Star Schema queries degraded in performance very quickly as the fact table grew, due to expensive hash joins on queries that matched larger number of dimension records. See chart below. We experimented with join indexes which did speed up the query, but prohibitively slowed down the data loads. John has all the supporting information and query plans�

So we modified the schema (and indices) and test queries to focus on a flat fact table structure, where facts and dimensions were pre-joined during loading, with the goal that hash joins could be avoided.

However, data loading times are problematic.. You can see from the chart below that load times increase with the size of the fact table. John helped me profile this behavior.. the elapsed time was almost all involved in the maintenance of two HG indexes applied to high cardinality name columns.. Disk I/O was pegged at 100% (iostats 杧) during these loads.. This implies that performance cannot be improved by further parallelization.. in fact the addition of additional parallel workloads can be presumed to compound the problem by competing for the same I/O resource.

Removing the HG indices dramatically speeds up the load times, but also dramatically slows down queries which filter on the name columns. LF indices are not an option as the name columns have more than 10,000 unique values.

So, that抯 the summary of findings.. Please let me know if you feel these findings are in any way inaccurate or unrepresentative of SybaseIQ 15_0 behavior, or if there are any solutions we haven抰 explored relating to the load I/O performance issues mentioned above.

NOTE: We tested Sybase 15.0 beta, understanding that it had improved parallelization of HG index updates, but unfortunately the bottleneck is disk I/O, not CPU.

Also, please let me know if you are aware of any upcoming product features that might improve the load performance of fact tables with HG indexes, specifically enhancements that might reduce the disk I/O associated with these updates and therefore remove or diminish the effects of I/O saturation noted above.

Detailed log:

I. 01/23 18:23:34. 0000025215 [20895]: Insert Pass 1 completed in 5 seconds.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C2_HG: Thread 7 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C2_HG: Thread 0 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C2_HG: Thread 2 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 4 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 0 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 5 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 2 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 3 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 6 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 1 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 7 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C2_HG: Thread 6 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C2_HG: Thread 4 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 DBA.fact_tmp.ASIQ_IDX_T733_C2_HG Sort Insert Statistics: 30 pages over 1 run(s) (Min: 30 Max: 30 Avg: 30) from 1 inserter(s).

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 4 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 5 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 0 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 2 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 7 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 6 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 3 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 1 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 5 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 4 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 0 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 7 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 6 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 1 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 3 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 2 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 5 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 4 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 0 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 7 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 6 Inserting 16370 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 [20024]: Index DBA.fact_tmp.ASIQ_IDX_T733_C1_HG: Thread 1 Inserting 16320 Recs, 0 Secs.

I. 01/23 18:23:35. 0000025215 DBA.fact_tmp.ASIQ_IDX_T733_C1_HG Sort Insert Statistics: 30 pages over 1 run(s) (Min: 30 Max: 30 Avg: 30) from 1 inserter(s).

I. 01/23 18:23:35. 0000025215 [20895]: Insert Pass 2 completed in 1 seconds.

I. 01/23 18:23:35. 0000025215 [20896]: Insert for 'fact_tmp' completed in 6 seconds. 491050 rows inserted.

I. 01/23 18:23:36. 0000025215 Cmt 85137

I. 01/23 18:23:36. 0000025215 PostCmt 0

I. 01/23 18:23:36. 0000025215 Cmt 85139

I. 01/23 18:23:36. 0000025215 PostCmt 0

I. 01/23 18:23:38. 0000025215 Insert Started.

I. 01/23 18:23:46. 0000025216 Cmt 85142

I. 01/23 18:23:46. 0000025216 PostCmt 0

I. 01/23 18:23:47. 0000025215 [20895]: Insert Pass 1 completed in 11 seconds.

I. 01/23 18:23:50. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 6 Inserting 16380 Recs, 2 Secs.

I. 01/23 18:23:51. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 4 Inserting 16380 Recs, 4 Secs.

I. 01/23 18:23:51. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 2 Inserting 16375 Recs, 4 Secs.

I. 01/23 18:23:52. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 3 Inserting 16380 Recs, 5 Secs.

I. 01/23 18:23:52. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 0 Inserting 16390 Recs, 5 Secs.

I. 01/23 18:24:00. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 5 Inserting 16305 Recs, 13 Secs.

I. 01/23 18:24:05. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 4 Inserting 16380 Recs, 18 Secs.

I. 01/23 18:24:05. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 1 Inserting 16380 Recs, 18 Secs.

I. 01/23 18:24:06. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 6 Inserting 16380 Recs, 19 Secs.

I. 01/23 18:24:06. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 5 Inserting 16380 Recs, 19 Secs.

I. 01/23 18:24:06. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 7 Inserting 16380 Recs, 18 Secs.

I. 01/23 18:24:06. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 2 Inserting 16380 Recs, 19 Secs.

I. 01/23 18:24:06. 0000025217 Cmt 85144

I. 01/23 18:24:06. 0000025217 PostCmt 0

I. 01/23 18:24:07. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 3 Inserting 16380 Recs, 20 Secs.

I. 01/23 18:24:07. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 0 Inserting 16380 Recs, 20 Secs.

I. 01/23 18:24:14. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 3 Inserting 16380 Recs, 22 Secs.

I. 01/23 18:24:17. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 7 Inserting 16360 Recs, 30 Secs.

I. 01/23 18:24:21. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 4 Inserting 16380 Recs, 16 Secs.

I. 01/23 18:24:22. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 6 Inserting 16380 Recs, 16 Secs.

I. 01/23 18:24:22. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 1 Inserting 16380 Recs, 16 Secs.

I. 01/23 18:24:23. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 5 Inserting 16380 Recs, 17 Secs.

I. 01/23 18:24:25. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 2 Inserting 16380 Recs, 19 Secs.

I. 01/23 18:24:25. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 3 Inserting 16360 Recs, 11 Secs.

I. 01/23 18:24:25. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 1 Inserting 16375 Recs, 38 Secs.

I. 01/23 18:24:25. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 7 Inserting 16380 Recs, 19 Secs.

I. 01/23 18:24:26. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 5 Inserting 16390 Recs, 26 Secs.

I. 01/23 18:24:26. 0000025218 Cmt 85146

I. 01/23 18:24:26. 0000025218 PostCmt 0

I. 01/23 18:24:26. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 7 Inserting 16315 Recs, 9 Secs.

I. 01/23 18:24:29. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 3 Inserting 16380 Recs, 22 Secs.

I. 01/23 18:24:29. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 0 Inserting 16380 Recs, 22 Secs.

I. 01/23 18:24:29. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 3 Inserting 16435 Recs, 4 Secs.

I. 01/23 18:24:37. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 1 Inserting 16340 Recs, 11 Secs.

I. 01/23 18:24:37. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 2 Inserting 16380 Recs, 46 Secs.

I. 01/23 18:24:37. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 5 Inserting 16435 Recs, 11 Secs.

I. 01/23 18:24:43. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 6 Inserting 16475 Recs, 53 Secs.

I. 01/23 18:24:46. 0000025219 Cmt 85148

I. 01/23 18:24:46. 0000025219 PostCmt 0

I. 01/23 18:24:47. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 4 Inserting 16380 Recs, 26 Secs.

I. 01/23 18:24:48. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 6 Inserting 16380 Recs, 26 Secs.

I. 01/23 18:24:48. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 1 Inserting 16380 Recs, 26 Secs.

I. 01/23 18:24:51. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 2 Inserting 16380 Recs, 26 Secs.

I. 01/23 18:24:51. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 5 Inserting 16380 Recs, 27 Secs.

I. 01/23 18:24:53. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 7 Inserting 16380 Recs, 27 Secs.

I. 01/23 18:24:53. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 6 Inserting 16380 Recs, 10 Secs.

I. 01/23 18:24:57. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 0 Inserting 16380 Recs, 28 Secs.

I. 01/23 18:24:58. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 3 Inserting 16380 Recs, 29 Secs.

I. 01/23 18:24:59. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 3 Inserting 16350 Recs, 30 Secs.

I. 01/23 18:25:02. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 0 Inserting 16380 Recs, 70 Secs.

I. 01/23 18:25:03. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 6 Inserting 16375 Recs, 9 Secs.

I. 01/23 18:25:04. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 0 Inserting 16350 Recs, 2 Secs.

I. 01/23 18:25:04. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 6 Inserting 16400 Recs, 1 Secs.

I. 01/23 18:25:06. 0000025220 Cmt 85150

I. 01/23 18:25:06. 0000025220 PostCmt 0

I. 01/23 18:25:08. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 3 Inserting 16375 Recs, 9 Secs.

I. 01/23 18:25:09. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 0 Inserting 16050 Recs, 5 Secs.

I. 01/23 18:25:12. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 4 Inserting 16380 Recs, 24 Secs.

I. 01/23 18:25:13. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 7 Inserting 16395 Recs, 47 Secs.

I. 01/23 18:25:13. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 6 Inserting 16380 Recs, 25 Secs.

I. 01/23 18:25:13. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 2 Inserting 16380 Recs, 22 Secs.

I. 01/23 18:25:14. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 1 Inserting 16380 Recs, 26 Secs.

I. 01/23 18:25:14. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 5 Inserting 16380 Recs, 23 Secs.

I. 01/23 18:25:14. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 4 Inserting 16370 Recs, 83 Secs.

I. 01/23 18:25:14. 0000025215 [20024]: Index DBA.fact.fact_dim_config_qualifiedifname_HG: Thread 7 Inserting 16030 Recs, 21 Secs.

I. 01/23 18:25:14. 0000025215 DBA.fact.fact_dim_config_qualifiedifname_HG Sort Insert Statistics: 120 pages over 1 run(s) (Min: 120 Max: 120 Avg: 120) from 1 inserter(s).

I. 01/23 18:25:15. 0000025215 [20024]: Index DBA.fact.fact_dim_time_datetime_min_HG: Thread 1 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:25:15. 0000025215 [20024]: Index DBA.fact.fact_dim_time_datetime_min_HG: Thread 4 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:25:15. 0000025215 [20024]: Index DBA.fact.fact_dim_time_datetime_min_HG: Thread 3 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:25:15. 0000025215 [20024]: Index DBA.fact.fact_dim_time_datetime_min_HG: Thread 5 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:25:15. 0000025215 [20024]: Index DBA.fact.fact_dim_time_datetime_min_HG: Thread 2 Inserting 98210 Recs, 0 Secs.

I. 01/23 18:25:15. 0000025215 DBA.fact.fact_dim_time_datetime_min_HG Sort Insert Statistics: 30 pages over 1 run(s) (Min: 30 Max: 30 Avg: 30) from 1 inserter(s).

I. 01/23 18:25:18. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 2 Inserting 16380 Recs, 41 Secs.

I. 01/23 18:25:19. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 5 Inserting 16390 Recs, 41 Secs.

I. 01/23 18:25:26. 0000025221 Cmt 85152

I. 01/23 18:25:26. 0000025221 PostCmt 0

I. 01/23 18:25:32. 0000025215 [20024]: Index DBA.fact.fact_dim_config_dev_name_HG: Thread 1 Inserting 16400 Recs, 55 Secs.

I. 01/23 18:25:32. 0000025215 DBA.fact.fact_dim_config_dev_name_HG Sort Insert Statistics: 90 pages over 1 run(s) (Min: 90 Max: 90 Avg: 90) from 1 inserter(s).

I. 01/23 18:25:32. 0000025215 [20895]: Insert Pass 2 completed in 105 seconds.

I. 01/23 18:25:32. 0000025215 [20896]: Insert for 'fact' completed in 114 seconds. 491050 rows inserted.

I. 01/23 18:25:40. 0000025215 Cmt 85153

I. 01/23 18:25:40. 0000025215 PostCmt 0

I. 01/23 18:25:41. 0000025215 Cmt 85155

I. 01/23 18:25:41. 0000025215 PostCmt 0

I. 01/23 18:25:41. 0000025215 Cmt 85157

I. 01/23 18:25:41. 0000025215 PostCmt 0

I. 01/23 18:25:41. 0000025215 Cmt 85159

I. 01/23 18:25:41. 0000025215 PostCmt 0

I. 01/23 18:25:46. 0000025222 Cmt 85161

I. 01/23 18:25:46. 0000025222 PostCmt 0

freelist fragmentation that was seen in 12.6. When running

workloads with a high number of truncates and deletes, this

aggressive reclamation caused more time to be spent in

searching for free space.

algorithm. It also refused to honor a join preference from the user requesting the use

of a pushdown join in such cases. This has been fixed.

- inserting into, or deleting from, or truncating the top table, or

- inserting into, or deleting from, or truncating the non-top table, and then

synchronizing the join index

may lead to incorrect results for subsequent queries using the join index.

This has been fixed.

earlier versions failed in 12.7 ESD #6 at opt_Select.cxx:1996.

LF_Max_Unique_Values

Default: 10000

Min: 1

Max: 200000

with Assertion 101506, if the user in question belonged to a

large number of groups.

new error code -1009416, rather than an internal error.

The error has been found and fixed per CR #461826

table with a SELECT of the same local temporary table

disappeared after a server sync.

and the emergency buffer pool is involved. This is a duplicate of CR 273582

and has been fixed.

join and large nested CASE statements was blank. Now the

output of the Join node contains the required entry.

Adding a file to a tempdb on a secondary node failed if the user wanted to store the raw partition on a secondary node.

ATOMIC block leaked server memory. Note that returning

a result set from an atomic block is not allowed and an

error is issued in this case; this behavior has not

changed.

dropped and the FORCE_DROP option was set.

caused a server failure.

became unresponsive waiting on the GetIdentityEnforceUniqueness

call.

<TableName>.<ColumnName>.substring(1,4) caused a server

failure. An error message is now returned.

If you try to execute a non existant stored procedure you correctly get this error.....

Msg 504, Level 11, State 0:

ASA Error -265: Procedure 'xxxxx' not found

If you try execute the non existant procedure but this time give it a parameter you should expect to just get the same error message say the procedure does not exist. However you get the following error instead....

Msg 201, Level 16, State 0:

ASA Error -615: Parameter '' not found in procedure '???'

This is misleading as it implies the procedure is OK but the parameter is incorrect.

As per the resolution in CR 570627.... IQ is coded to create the extract output files with permissions rw-r--r--. umask acts on the permissions as follows:

When a file is created, its permissions are set to what the creating program asks for minus what the umask setting forbids. So, the bottom line is that umask can't add any permissions that the creating program hasn't given.

We need a FR that allow control.

1. Fail with a signal 11. The IQ server continues to run.

Only an incomplete stacktrace is generated.

2. Fail with signal 5 or 11. This may be followed by a message

"Memory corruption detected", another signal 11 (in which case the server continues to

run with no complete stacktrace generated), or by multiple signal 5's (in which case the

server fails with a full stacktrace generated).

This has been fixed.

AIX -

Feature request -

When starting up the dbremote via SCJ, allow to user define an unique dbremote log file name to be used in all startup times, instead of to rename the Dbremote log file name using the string like this :

engine_name.xxx.dblog

where xxx is incremented each time that DBremote is started .

When starting dbremote via script, the unique dbremote log file name is defined used -o switch .

when a table, that existed in 12.6, is opened for the first time in 12.7.

A feature was added to 12.7 to construct a null bitmap on each

nullable column defined on the table. This bitmap is constructed

when the table is opened readwrite for the first time on 12.7 after

the upgrade has been done. Subsequent readwrite operations on the table

will not incur this cost.

The cost is a function of the number of NULL and non-NULL cells

on a given column in conjunction with the number of rows in the table.

This null bitmap was added, to address performance issues,

to greatly improve query performance when a query needed

to take into account NULL values.

in different case, was allowed after the server was restarted

in CASE RESPECT database.

function is 255. If the precision of the numeric function exceeds the maximum

value supported in SQL Anywhere, the following error occurs: "The result

datatype for function '_funcname' exceeds the maximum supported numeric

precision of 255. Please set the proper value for precision in numeric function,

'location'"

15.0.0 ESD#1 version string. This release will incorrectly report its

version as 15.0.1 ESD#1. This issue has no effect on the functioning of the

software.

IQ dummy, the server failed. An appropriate NULL check was

added in the function to avoid a server failure.

loop

savepoint;

fetch next from cursor;

update table where current of cursor;

end loop

predicates, query performance degraded when using 12.7

instead of 12.6.

Even when there are no users connected to the IQ reader servers, there are internal event connections happening. To verify this, disable ev_iqmpxq2w on the IQ reader servers and ensure there are no user connections. Run the ddl on IQ writer server and observe the IQ reader server is consuming same memory as the IQ writer.

Once the TLV replay is completed, the IQ reader server releases the extra memory that it consumes while managing the versions. However, on UNIX, that memory is never returned to the operating system. Once memory is allocated for a process, the process retains that memory until the operating system is restarted.

To avoid having the memory consumed and not released, the indexes must be grouped in BEGIN PARALLEL IQ and END PARALLEL IQ. See "The CREATE INDEX statement" in Chapter 6, "Using Sybase IQ Indexes" in System Administration Guide: Volume 1 and "BEGIN PARALLEL IQ ... END PARALLEL IQ statment" in Chapter 1 "SQL Statements" in Reference: Statements and Options.

cast the select list as appropriate. For instance, in the following case:

create table test (c1 decimal(15,2), c2 decimal(18,2));

select abs(c1) + c1 + c1 + c2 into #tmp from test;

change the select statement to

select cast((abs(c1) + c1 + c1 + c2) as decimal(15,2)) into #tmp from test;

tables, and if the query contained many tables, then there

was a chance the server would give assertion 101508, instead

of giving the "syntactic limit exceeded" error.

expressions are used in 12.7 code. To get better

performance for queries containing complex expressions in

predicates, set following internal option: set

[temporary] option dml_options6 = 32768 Note: Some queries

might show performance degradation after setting this

option. This option should be set to get better

performance for queries which do not show better

performance by default.

while rewriting a very large and complex SELECT statement.

This happened if the engine was close to running out of

stack size or was low on cache size. The problem occurred

with queries containing UNION, EXCEPT, INTERSECT, or deeply

nested sub-selects. The engine now returns the error "-890:

Statement size or complexity exceeds server limits."

LOAD aborts the engine during 2 to 3 byte FP rollover while destroying a locked buffer.

unresponsive server. Now an error is reported.

For a given year, month, and day:

year = current_year - 1;

days_in_year_0000 = 366;

binaryDateValue = (year * 365)

+ (year / 4)

- (year / 100)

+ (year / 400)

days_in_year_0000

day_of_current_year

-1;

It should read

For a given year, month, and day:

year = current_year - 1;

days_in_year_0000 = 366;

binaryDateValue = (year * 365)

+ (year / 4)

- (year / 100)

+ (year / 400)

+ days_in_year_0000

+ day_of_current_year

-1;

create table t1(num as integer, name as varchar(30))

...

case

when (select count(*) from sa_conn_properties(),t1) > 1

then ...

else ...

end case

...

Will cause the server to crash. To prevent the crash, declare t1 IN SYSTEM or execute the predicate query outside the case statement and assign it to a variable, then use the variable in the case predicate.

If you download thuis file and use tar -xzvf to unzip it, then run sybinstall. The jConnect 6.0 part throws tar an gzip erros (see below). The sybiqntwkclnt127lin.tgz file from SPDC download site for the same software is fine and works OK.

-----------------------

Installing jConnect 6.0

-----------------------

jConnect-6_0/

jConnect-6_0/index.html

jConnect-6_0/netimpct.gif

jConnect-6_0/version.txt

jConnect-6_0/tools/

jConnect-6_0/tools/decode-tli

jConnect-6_0/sp/

jConnect-6_0/sp/sql_asa.sql

jConnect-6_0/sp/sql_server.sql

jConnect-6_0/sp/sql_server12.5.sql

jConnect-6_0/sp/sql_server12.sql

tar: Skipping to next header

tar: Archive contains obsolescent base-64 headers

gzip: stdin: invalid compressed data--crc error

gzip: stdin: invalid compressed data--length error

tar: Error exit delayed from previous errors

Install failed for the following reasons

----------------------------------------

Copying of jConnect 6.0 failed

tar: Skipping to next header

tar: Archive contains obsolescent base-64 headers

gzip: stdin: invalid compressed data--crc error

gzip: stdin: invalid compressed data--length error

tar: Error exit delayed from previous errors

collisions occurred in some situations.

use of ROWID, and if the table specified to ROWID was a

remote table, and if the query was processed in partial

passthru, then there was a chance the server would fail.

OFF prior to the creation of a table sometimes affected the

result of calculations involving the numeric columns of the

table.

few rows. This happens if the engine has found redundant conjuncts and try to remote it.

An example of this is the query

select 1 from T

where a = 1 and ( b = 2 or c = 8 ) and ( d = 4 or e = 10 ) and ( a = 1 or e = 7 or c = 9 )

This has been fixed.

the server became unresponsive after detecting a stack

overflow.

to a large number of groups. This has been fixed.

are no longer released with any ESD beyond ESD #8. Customers

who did not install 12.7 ESD #8 will be required to install

ESD #8 to update to the latest supported version, before

any newer update can be applied.

Central may fail with an error about the path to the shared IQ store not existing on

the secondary node, if the path to the database file (the .db file) contains Chinese

or Japanese characters. There are two workarounds:

1. Restart the Syabse IQ Agent service using the user account that was used to log

on to the IQ server host. In the list of services, right click Sybase IQ Agent

15.1 -> Properties -> the "Log On" tab -> This account, and enter the user account

id, the password, confirm the password, and click the OK button. Click the OK button

if a box pops up with the message "The new logon name will not take effect until you

stop and restart the service", and then restart the service by right clicking Sybase IQ

Agent 15.1 in the service list and selecting Restart. Or if the service is not already

running, start it by right clicking Sybase IQ Agent 15.1 in the service list and

selecting Start.

Or

2. Stop the Sybase IQ Agent service and start the Sybase IQ Agent as an application:

In a command window:

SybaseIQagent15.exe -r "SybaseIQAgent15"

- run GUI install

- run standalone GUI install; setup from sysam-2_10 dir.

- run standalone Silent install from sysam-2_10 dir. (see the command below)

note: change the "$V(SybRoot_InstallLocation)" below to be the path to the root of the install location (ie. $SYBASE), and change "suite_sysam_sysam_server.active=true" value to be true or false depending if you want to install the sysam server or not and change "suite_sysam_sysam_util.active=true" value to be true or false depending if you want to install the sysam utilities or not

--------------

./setup -silent -W SybaseLicense.agreeToLicense=true -W setupTypes.selectedSetupTypeID= -P suite_sysam_product_root.active=true -P installLocation=$V(SybRoot_InstallLocation) -P suite_sysam_sysam_server.active=true -P suite_sysam_sysam_util.active=true

In IQ 12.5 and 12.6, conditions involving a simple functional or arithmetic

expression over a single column could be evaluated using a scan of

the FP index. For example, the condition "SQRT(T.X) > 10" could be

evaluated in this manner.

In IQ 12.7 the optimizer and query run-time system were enhanced to

enable any arbitrarily complex condition over a single column, such as

"SQRT(T.X) > (10 + (100 * LN(T.X)))" to be evaluated in the

same manner.

In the vast majority of queries, evaluation of complex conditions using

this optimization results in improved overall query performance. However,

as with most optimizations, there are rare exceptions. This change

provides a mechanism that allows an IQ server to revert its behavior for

this optimization back to applying only to the simpler conditions supported

in IQ 12.6.

load table (f1 prefix 1, f2 binary with null byte,

f3 binary with null byte) from 'test-iq.001'

format binary quotes off;

are not prohibited by the syntax and may load misaligned data to the table.

testing of data, which could them be made "live" by simply changing the owner.

"I. 09/11 08:47:35. 0000000171 O/S Err#: 0, ErrID: 9216 (df_Exception); SQLCode: -1009113, SQLState: 'QSB13', Severity: 14

I. 09/11 08:47:35. 0000000171 [20134]: Secondary error noted"

"secondary error noted" message text has been clarified.

I. 09/09 16:36:31. MPX dynamic collision on dbo.billed_roc -- connection 16975

I. 09/09 16:36:31. User "dbo" dropped connection 16975 ("vkatya")

I. 09/09 16:36:32. -- dropped 1 of 2 connection(s) for table: dbo.billed_roc -- but 1 connection(s) are still holding a lock

I. 09/09 16:36:32. -- Too many retries on dropping user connections. Server is shutting down.

crash. This has been fixed.

certain Japanese characters. In this case, the Interactive SQL Classic (dbisqlc.exe) or

Interactive SQL Java (dbisql.exe) tool can be used instead to create the database. If

these tools are started in the command line and a connection string (-c "...") is

provided., ensure either to avoid using the links parameter, or if the links parameter

is provided, to use parentheses instead of curly braces around the host name and the port

number, e.g.,

Instead of

dbisql -c "uid=dba;pwd=sql;eng=mysrv;links=tcpip{host=<host name in Japanese>;port=8888}"

use either

dbisql -c "uid=dba;pwd=sql;eng=mysrv"

or

dbisql -c "uid=dba;pwd=sql;eng=mysrv;links=tcpip(host=<host name in Japanese>;port=8888)"

I. 10/01 17:53:06. 0000030825 O/S Err#: 0, ErrID: 9216 (df_Exception); SQLCode: -1005028, SQLState: 'QTA28', Seve

rity: 14

I. 10/01 17:53:06. 0000030825 [21092]: Syntax error - RANK/DENSE_RANK/PERCENT_RANK/NTILE/CUME_DIST function ORDER BY

item must be expression over column.

-- (df_Heap.cxx 3572)

I. 10/01 17:53:06. 0000030825 Memory corruption detected

I. 10/01 17:53:06. 0000030825 Idx: 0 Addr: 0x2aaace5c2550 Size: 16 File: df_Heap.cxx Line: 3533

I. 10/01 17:53:06. 0000030825 Usertype: df_Heap::GenerateAggregate _fence bad

may cause error "Cannot evaluate expression". For example,

SELECT r3.id, Col1 = r3.id

FROM DBA.fin_code r7 cross join DBA.customer r3

GROUP BY ROLLUP (r3.id )

lines at the beginning of the file. For details on the

HEADER SKIP clause, see the Sybase IQ 15.2 Release

Bulletins.

SQL Anywhere Server - Database Administration 粻Starting and Connecting to Your Database 粻The database server 粻Database server options 粻-n server option

The workaround is to enter other connection information but NOT the IQ server name.

multiple machine instructions, due to a compiler flag

+u4. This led to a race condition resulting in a server

failure. Sybase IQ now has double byte aligned structures

and has removed the use of +u4, so that the pointer

assignments are atomic. This modification prevents any

race condition on pointer assignment and failures due

to this race condition.

Example:

create table TestDensRank_desc(col1 int,col2 int, col3 int);

select

col1,

col2,

xyz=1,

max(col3),

day_order = dense_rank() over( order by col1,col2, xyz desc )

from TestDensRank_desc

group by

col1,

col2;

SQL Anywhere Error -1003002: CtLibrary Error: 44, Severity: 0, Origin: 1, Layer:

4

Error Message: ct_connect(): protocol specific layer: external error: The

attempt to connect to the server failed.

OS Error: 0, OS Message: ---.

--

(hqm_con

(0 rows affected)

dbisql -c "dsn=Rey" -nogui

Could not connect to the database.

Data source name not found, and no default driver specified

SQLCODE=0, ODBC 3 State="IM002"

e.g. Instead of

start_iq -n <chinese_server_name>

please use:

start_iq -n "<chinese_server_name>"

In IQ 15.x

sp_iqpassword and Alter User / Create User can not contain the following chars for the password additional to the ones of 12.7.

And the Password can not start with a numeric Character.

Words used as reserved words, like "select" or "as" neither can be used any more.

IQ Database was setup with Charset 1252

can be used:

char ASCII

# 35

$ 36

0 ... 9 if it is not first character

@ 64

A 65

....

Z 90

_ 95

a 97

....

z 122

€ 128

� 140

...

higher then ASCII 140 some chars can be used some not.

Sybase IQ Unload Utility Version 15.1.0.5034

Connecting and initializing

***** SQL error: Invalid database server command line

load.

The problem can happen with partitioned or non-partitioned

tables and variable or fixed width loads.

alter, update, delete, references) on a table, loading

a table resulted in the error "Permission denied: you

do not have permission to select from "<table_name>"

SQLCODE=-121, ODBC 3 State="42000". The correct behavior

requires only select, insert, and alter permissions to

perform a load.

This has been fixed.

However, the input parameter for sp_iqindexsize can easily exceed 128 and cause procedure to fail.

Here is a real life example, of proc failing:

sp_iqindexsize('price.StressPrices_prices_Clone0_46668739_nyrmdla205_ny_fw_gs_com_574_20091030.[i_StressPrices_prices_Clone0_46668739_nyrmdla205_ny_fw_gs_com_574_20091030_3_HG]')

[Sybase][ODBC Driver][SQL Anywhere]Invalid user ID or password

Error code=-103

SQL state=28000

Msg 7343, Level 16, State 2, Line 1

The OLE DB provider "SAOLEDB.11" for linked server "SYB_IQ" could not INSERT INTO table "[SYB_IQ]..[DBA].[emp1]".

sp_iqconnection stored procedure. This has been fixed.

server to abort. This has been fixed.

have permission to grant permissions to "<username>".

Permission is denied trying to add a group member from a group login ID.

You can use Syntax 1 or 2 of the GRANT statement if at least one of these conditions

is met:

1. You are changing your own password using GRANT CONNECT

or

2. You have DBA authority

This is 127 ESD 7 on linux power pc 64 bits (but problem reproduceble on linux intel 64).

Take the example query " select distinct(substring (char_column,1,1)) from table1 " under following conditions:

- table1 has several million rows

- Column char_column has a low cardinality and an LF index.

IQ resolved the query applying substring function to all rows and then applying the distinct.

If it would resolve the query calculating the distinct using the index first and then applying substring

then the performance improves a lot. These are examples sent by customer:

select distinct(char_column) from table1 -> 0.063 (seconds LF index is used)

select distinct(substring (char_column,1,1)) from table1 -> 76 seconds (index is not used).

select substring (A,1,1) from ( select distinct(char_column) A from table1 ) t -> 0.063 seconds.

This has also been tried in 15.1 and in this case performance for second query improves a lot comparing to 12.7, as IQ 15.1 uses palalelism and performance is better but not as good and doing first the distinct and second the subsctring.

such as:

1. home <-, or End ->, up, down key

2. recognise some special key combo ( like ALT+< ) to go back to previous command

an IQ server failure.

period of time, when a long running connection was

disconnecting.

http://www-01.ibm.com/support/docview.wss?rs=32&uid=swg21110831

Downloaded and install "vacpp.80.aix53TL6-61.oct2009.ptf.ta" from here:

ftp://ftp.software.ibm.com/aix/products/ccpp/ptfs/vacpp.80.aix53TL6-61.oct2009.ptf.tar.Z

Msg 21, Level 14, State 0:

SQL Anywhere Error -1006256: Invalid arithmetic construction (1)

--

(hos_numeric.cxx 6844)

The circumstances are as follows:

An IQ table is defined as "create table local_t1 ( c1 numeric(19,4) );"

A remote table is defined as "create table remote_t1 ( c1 numeric(19,4) );"

The statement producing the error:

"insert local_t1 location 'remote.db' { select sum(c1) from remote_t1 };"

The remote server will calculate the sum and likely require more storage than is needed for a numeric(19,4). This will cause a buffer in IQ to be written incorrectly.

Now that this correction has been made, a different error may be reported when a numeric data type cannot be converted to a numeric data type of a smaller size.

We may now report error: "SQLCODE -1001006, Conversion is not possible" instead of

"SQLCODE -1001010, Numeric value out of range".

xp_cmdshell procedure on AIX and Linux platforms (already

implemented on Sun).

example, by pressing CTRL-C in the console to which the

server is attached) sometimes caused a server failure.

the MPX to crash:

1. dba creates a stored procedure or a table specifying dbo as the owner

2. dba drops the stored procedure or the table without specifying the owner

3. dba re-creates the stored procedure or the table as in step 1

This has been fixed.

an incorrect result.

select * from saoledblink..SYS.syscollation;

would have failed with the error:

Msg 7356, Level 16, State 1, Line 1

The OLE DB provider "SAOLEDB.11" for linked server "saoledblink" supplied nconsistent metadata for a column. The column "collation_order" (compile-time ordinal 4) of object ""SYS"."syscollation"" was reported to have a "DBCOLUMNFLAGS_ISFIXEDLENGTH" of 0 at compile time and 16 at run time.

The problem occurred for BINARY(n) columns. The DBCOLUMNFLAGS_ISFIXEDLENGTH schema attribute was set true (0x10) at run time. This problem has been fixed.

a multiplex coordinator requires the utility database to be

started with the same server name. This is a change in

behavior since 12.7 and was also a problem for users

attempting to restore, when the production server was running

using the coordinator server name.

This has been fixed. The message is now "Host variable 'xxx' is not initialized"

where xxx is the name of the uninitialized parameter.

Ja_JP

Ja_JP.IBM-932

the temporary cache is running low. Please contact Sybase Technical Support for

assistance, if this issue occurs.

update one column of each row with the CURRENT OF clause.

All is fine.

But, in the cursor fetch loop, when we add a select count (*) from another table,

then IQ SILENTLY updates a column that is not specified in the update statement!

join, while executing a query with many predicates

between multiple tables, a non-fatal error sometimes

occurred.

the "IQ Internal error. Please report this to Sybase IQ support. -- (df_Heap.cxx 2695) "

error. This has been fixed.

this case numeric(2) and numeric(5)) a wrong result may occur in the case of HPD Join. This has been fixed.

Primary key for "ISYSIDX" has missing index entries

Foreign key "ISYSTAB" has missing index entries

Index name "index_name" has missing index entries

perform poorly in some cases.

a byte FP lookup table inconsistency error.

*** ERROR *** Assertion failed: 201501

Page 0x0:0x0 for requested record not a table page

UpdCur not updatable in IQ

Sample Statement (missing from clause in select statement in where section):

delete Contacts_new

where surname in ( select distinct surname Contacts

union

select distinct surname Contacts

)

Now the above kind of query will not lead to crash. If Contacts_new and Contacts will

1) IQ table- We will get the error "Update operation attempted on non-updatable remote query".

The able mentioned query , internally converted to the following

delete Contacts_new

where surname in ( select distinct surname Contacts

union

select distinct surname as Contacts from SYS.DUMMY

)

SYS.Dummy is the SA table thatswhy we will get this error.

2)SA table - All the rows of Contacts_new will be deleted or Contacts_new table will truncated. Because , Here column "surname" is not found in the subquery, then it will be search in the outer query block. Since table "Contacts_new" has a column "surname" it is a valid out reference and the WHERE clause is always true.

a multiplex node failure at saint_iqthresholdddl.cxx:6232 with

the error "Unexpected error during commit."

OpenServer application to 32k.

large size backups of Sybase IQ 15.x databases, compared

to backups of 12.7 databases.

the server was restarted after it stopped for a disk-full

condition on the disk where the transaction log resided.

conjunctive condition cause the query to fail with non-fatal stack trace at

opt_TableNode.cxx.

values for the DbspaceSize column, when dbspace sizes were

over 4GB.

in versions 12.7 and 15.1, @@rowcount behaves as expected.

AVG) with COALESCE(isnull, ifnull) returned NULL, when the

query optimizer chose parallel GBH.

temp buffer cache, if all pages of the cache were locked or

dirty. This error was reported: "main Bufman: All buffer

cache pages are in use, ask your DBA to increase the size

of the buffer cache."

State 0: SQL Anywhere Error ~1006167: DateTime is invalid

(0)".

opt_Driver.cxx:5700" was reported by an IQ error checking

routine to ensure that the IQ server did not fail due to

invalid data structures. The routine checked whether a column

from a view or derived table was valid for a partitioned table

optimization, where IQ can treat the output of a UNION ALL

as a single table. The check was modified to permit the

valid query.

query server date_order did not propagate to the query

server. View creation now propagates to query nodes. After

creating a view using date/timestamp/datetime as filters,

set the date_order option on the query node to the same

value as the date_order option on the writer node to

perform a query on that view.

s_bm.cxx:2807 and the error "O/S Err#: 0, ErrID: 2049

(sbm_exception); SQLCode: -1009045, SQLState: 'QSA45',

Severity: 14 Bitmap failed internal checks. You may have

a damaged index. Please check your IQ message file for

additional information." and rolled back.

for example, ORDER BY ROWID(table) DESC, returned

incorrect results.

have caused a server failure with inconsistent memory

detected.

partition on AIX sometimes caused the server to fail in

s_dbmvbio::PlaceNewSegment.

created with collation 950ZHO_TW to import data from a

remote ASE server generated an exception thrown from

hqm_con.cxx:182 and reported the error "SQLCode: -1006000,

SQLState: 'QBA00', Severity: 23 IQ Internal error. Please

report this to Sybase IQ support."

a pending status. A timing sensitive deadlock could occur

in the IQ main or temp buffer cache, if all pages of the

cache were locked or dirty.

This option must be set: set option Date_First_Day_Of_Week = 1;

server failure in s_dpThreeByteFetch::GetCardinalty or inconsistencies in

memory.

to a windows form resulted in an error. Visual Studio

could not generate DataSets and Entity Framework models

using Sybase IQ data sources.

menu returned the error "The System could not find the

file specified." Visual Studio was missing entries for

Interactive SQL and Sybase Central.

DSN did not list any configured DSNs. The Visual Studio

integration component now searches Sybase IQ data sources

using the prefix "SYBASE IQ".

clause must be expression over column" is generated, if the

ORDER BY of the WITHIN GROUP clause of the PERCENTILE_DISC

function is over a constant expression. This error

is correct, since the ORDER BY over a constant expression

makes no sense. The error message is now more accurate.

contained a select list item that was not a table

column could cause a server fault, when executing in

an IQ table or a proxy table. The error reported was

"SQLCode: -1001047, SQLState: 'QFA43', Severity: 14

Column: is not accessible in this context. --

(df_GenerateDF.cxx 192)."

optimization that affected push-down join conditions.

A list of IMSL errors are in files imslerr.dat and imsls_e.dat.

A copy of those files can be opened in any text editor.

contained CASE statements failed with the error

"IQ Internal error. Please report this to Sybase IQ

support. -- (dfjo.cxx 198) [-1006000] ['QBA00']."

went out of scope, the server failed at s_blockmap.cxx:2440.

of time some of the writers shut down with a non-fatal

IQ internal error detected at s_bm.cxx:1476 and the

message "Bitmap result bitcount not in legal range."

IQ fails in the dirty page to a buffer cache at creation.

an incorrect result. The result depended on the method the query

optimizer chose to process the pushed down predicate of a hash

pushed down join.

Please document that a list of errors can be found in the files imslerr.dat and imsls_e.dat

imsl library functions then only errornumber and no errortext

is put into the logfile or send to the client

/opt/sybase/IQ-15_1/IQ-15_1/lib64/libiqserv15_r.so [0x400011f52d8]

/opt/sybase/IQ-15_1/IQ-15_1/lib64/libiqserv15_r.so [0x400011f5534]

/opt/sybase/IQ-15_1/IQ-15_1/lib64/libiqserv15_r.so [0x400011d6b88]

There is a seperate CR 633385 raised to resolve this stack print error on IBM platforms.

The real stack can be obtained from running s gbd on the core, failure is at _ZN15df_MemGovRegion14DoCanAllocPageEv

I. 05/18 09:29:13. TDS 178: login: user 'DLNPSPSW04_IQ_maint', database '', application 'RepServer', host 'lns66d-8110a', host_pid '21988', charset 'iso_1'

I. 05/18 09:29:13. Login procedure 'DBA.sp_iq_process_login' caused SQLSTATE '22003'

I. 05/18 09:29:13. Value 2117418877184 out of range for destination

The login is not rejected however and the connection to IQ is allowed (this must surely also be a bug?). Another side affect of the error is that the call to sp_login_environment() is not made meaning that the sp_tsql_environment procedure is not called so database options specific to TDS connections are not set.

The source of the "Value 2117418877184 out of range for destination" error is an incorrect declaration of the column that represents LockName in the stored procedure sp_iq_process_login...

'create procedure DBA.sp_iq_process_login()\x0Abegin

declare @admin_on char(1);

declare @has_perm char(1);

declare @login_locked char(1);

declare @number_connects integer;

declare @number_connects_admin integer;

declare @number_db_connects integer;

declare @password_days integer;

declare @password_days_admin integer;

declare @password_created timestamp;

declare @times_connected integer;

declare c1 integer;

declare c2 varchar(255);

declare c3 varchar(255);

declare c4 integer;

declare c5 varchar(255);

declare c6 varchar(255);

declare c7 integer;

declare c8 varchar(255);

declare c9 varchar(255);

declare c10 varchar(255);

declare c11 integer;

declare c12 integer;

declare c13 integer;

declare c14 unsigned integer;

declare c15 integer;

declare @counter integer;

declare end_of_cursor exception for sqlstate value ''02000'';

declare saconn dynamic scroll cursor for call sa_conn_info();

......

the declare for c14 (which is the LockName) should be unsigned bigint and NOT unsigned integer.

Here is the corresponding declares in sa_conn_info where the connection infromation is obtained from...

'create procedure dbo.sa_conn_info(in connidparm integer default null)

result(Number integer,Name varchar(255),Userid varchar(255),DBNumber integer,LastReqTime varchar(255),ProcessTime varchar(255),Port integer,ReqType varchar(255),CommLink varchar(255),NodeAddr varchar(255),LastIdle integer,CurrTaskSw integer,BlockedOn integer,LockName unsigned bigint,UncmtOps integer)

begin

declare connid integer;

declare local temporary table t_conn_info(

Number integer null,

Name varchar(255) null,

Userid varchar(255) null,

DBNumber integer null,

LastReqTime varchar(255) null,

ProcessTime varchar(255) null,

Port integer null,

ReqType varchar(255) null,

CommLink varchar(255) null,

NodeAddr varchar(255) null,

LastIdle integer null,

CurrTaskSw integer null,

BlockedOn integer null,

LockName unsigned bigint null,

UncmtOps integer null,

)

in SYSTEM not transactional;if(connidparm is not null) then

....

Not an issue for IQ 15 and higher as this use dbo.sp_login_environment instead of DBA.sp_iq_process_login

when using MS-SQL SSIS. Could not insert, update, or delete,

when using MS-SQL 2008 Link Server to Sybase IQ. The error

"[ADO NET Destination [16]] Error: Length cannot be less than

zero." was returned. ADO.NET drivers were not properly registered

with the system registry for Microsoft Visual Studio 2008.

the cursor_window_rows option and caused the server to become

unresponsive.

no data was sent to Sybase IQ and the connection was dropped

abruptly. The schema definitions that provide input to the

OLEDB driver have been updated to return correct database

metadata for Sybase IQ.

could cause a server failure, if the specified column did not

exist. A server failure could also occur, when calling stored

procedures that operate on literal names, when quotes were

omitted. For example, sp_helptext dbo.sp_iqstatus

caused a server failure, because without quotes,

dbo.sp_iqstatus was interpreted as a table/column name pair.

The client for which Japanese included in the result of SQL is using which character set is also converted correct.

: ===== Thread Number 1305012544 (IQ connID: 0000514253) =====

/sybase/IQ-15_3/bin64/../lib64/libiq15.so(_Z9pcstkwalkP9stk_traceiP6db_logP6hos_fd+0x31) [0x2aaabf3a4ad1]

&/sybase/IQ-15_3/bin64/../lib64/libiq15.so [0x2aaabf3a3a42]

&/sybase/IQ-15_3/bin64/../lib64/libiq15.so(_Z14DumpAllThreadsPKcji+0xff) [0x2aaabf3a37bf]

&/sybase/IQ-15_3/bin64/../lib64/libiq15.so(_Z9hos_ABORTPKcjS0_PcS1_+0x19a) [0x2aaabf15729a]

&/sybase/IQ-15_3/bin64/../lib64/libiq15.so(SigHndlr+0x4e) [0x2aaabfc6366e]

&/lib64/libpthread.so.0 [0x367520eb10]

&/sybase/IQ-15_3/lib64/libiqserv15_r.so [0x2ad89f7f21e7]

&/sybase/IQ-15_3/lib64/libiqserv15_r.so [0x2ad89f5f9e00]

&/sybase/IQ-15_3/lib64/libiqserv15_r.so [0x2ad89f5fad8f]

&/sybase/IQ-15_3/lib64/libiqserv15_r.so [0x2ad89f8eb42f]

&/sybase/IQ-15_3/lib64/libiqserv15_r.so [0x2ad89f8c314f]

&/sybase/IQ-15_3/lib64/libiqserv15_r.so [0x2ad89f8c3313]