Sybase数据库技术，数据库恢复专家

with 11" in the module 'des_rekeep' together with a stack trace

which includes the modules 'proc__linktopss' and 'lwp_proc_setup'

may be reported in the error log when a client application is

executing a query using "dynamic sql" and the system is configured

with a low value for the configuration parameter "number of

open objects" according to its workload.

and a temporary table has at least one BIT column.

allocation at the same time. The deadlock information will consistently show the deadlock

ccurring on rowid 0 for both spids.

A high value for 揵y Log markers� indicates that Adaptive Server is flushing the ULCs because of

permanent log marker scans (a permanent log marker scan indicate that syslogs includes the log

record). Adaptive Server uses the log records for operations like executing triggers, rollbacks,

aborts, and so on. When Adaptive Server needs a permanent log marker but does not have one, it

flushes the ULC to create a new log marker. When 揵y Log markers� displays a large percentage of ULC

flushes, you may need to reduce the number of unnecessary or redundant triggers, rollbacks, or aborts.

creates a table and a procedure referring to that table, then

execution of that procedure following a schema change such as

dropping and recreating the table will result in an "object

not found" error (Msg 208).

by memory corruptions or abnormal process termination when the optimized

ASE binary is in use. This problem is most likely to manifest on Linux 64-bit

architectures.

for queries involving distinct multi-table joins.

when LDAP User Authentication is enabled (configuration option 'enable ldap user auth' is set to

1 or 2) and ASE is performing multiple user authentication simultaneously.

This bug can be resolved by installing patch for Solaris BugID 6972029

(contact Oracle for the OS patch, sun sparc: 144190-03, sun x64 144191-02).

Alternatively, it can be resolved in ASE through a workaround fix by enabling boot time trace flag 9415.

may be reported during boot recovery.

may fail to execute correctly producing stack traces.

because they appear to be errors. The messages will no longer be printed by sp_dboption.

even when the spid in the command is valid.

while booting up the higher version server to apply upgrade, ASE will raise 110272 error.

(a primary key and foreign keys) with each other.

A more complete version of this fix has been developeed under CR 614823, i.e. 614823 supersedes 588137.

monInMemoryStorage

monPCISlots

monPCIEngine

monPCIBridge

monProcedureCacheModuleUsage

monProcedureCacheMemoryUsage

cache size, cache creation will be failed.

the Java associated thread executing on the behalf of the ASE task will not be

notified to clean up its resources. This will result in leaking thread and memory

resources at operating system level.

it will issue the Replication Command Language command 'resetqueue' by default

when starting up against a database with a durability value of 'no_recovery',

instead of incrementing the primary database generation ID, so to prevent

the likelihood of hitting a limit while updating the generation ID value.

The current ASE fix will be effective when the Replication Server CR 615322

is also released in some later Replication Server 15.5 ESD version.

refer to ASE CR 230301 and refer to RepServer CR 370359 ... customer wants a feature request added where the RS thread is never the deadlock victim on a replicate db. He's had issues in the past where even after multiple retries, the tran is still in a deadlock causing the DSI to suspend.

If the rs thread is chosen for the deadlock victim, in some cases even after attempting the tran multiple times it still deadlocks, causing the tran to not complete and the DSI to suspend or if set to ignore the primary and replicate dbs become out of sync

(as seen in showplan output) and at least one of the Scan Operators under the Join is

doing dynamic partition elimination (also shown in showplan), the query may return too few

rows. This problem ocurrs in queries with at least three tables and which contain join

predicates involving columns of the partitioned table that are not part of the partition

condition of that table.

This may be followed by SIGSEGV, SIGBUS or timeslice error and lead to ASE hang.

ASE reports incorrect name of the function from which it is raised.

terminated for query involving text column, nested convert with concatenation to correlated

sub-query result.

and miss the first sample when the timestamp of sample equal to the start time of playback session

the given database was previously quiesced.

with the stack trace including modules 'NormalizeScalar' and 'Optimize'

when executing a SELECT statement with a GROUP BY clause which joins

two or more views including subquery(s) in the SELECT list of the view(s).

The query must be a non-ANSI extended column aggregate query, which

references columns outside the GROUP BY list, in the SELECT list or

HAVING clause, outside aggregate expressions.

dbcc table_xfer_check.

being stored in a database table.

approaching the maximum allowed) always be sent to the errorlog when it is raised.

of the cluster when it is changed on one of the nodes and the table is not marked explicitly

for replication.

00:00000:00066:2009/04/01 13:44:09.02 server RepAgent(41): Connecting to a Replication Server that supports a lower LTL version. Features that are not supported by the lower LTL version may cause the RepAgent to shutdown. Set the 'skip unsupported features' configuration option to prevent shutdown.

00:00000:00066:2009/04/01 13:44:09.02 server RepAgent(41): The Replication Server LTL version should be at least 710 for full support of features. (Replication server LTL version 700, Rep Agent LTL version 710).

a 6901 error, "Overflow on High component of timestamp occurred in database

<dbid>. Database table possibly corrupt". This is no longer possible.

After warnings that the database timestamp is reaching its limit, the

database will finally be shut down, leaving the database in read-only mode.

there is an expression in the select-list that requires a sequencer

node to evaluate it(e.g. - any single table column that appears in an

expression in the select-list). This will occur in parallel queries

if an XCHG op is placed between sequencer node and union node where

it must be evaluated.

the following error message occurs:

Error while executing SQL for migration of 'LK.dbo..au_pix'.

Root Cause:Msg 17741, Level 16, State 1

Server target, Procedure sp_primarykey, Line 131, Status 0, TranState 1

Primary key already exists on table -- drop key first.

Recoverable error. Task 'LK.dbo..au_pix' for path 'pubs2 --> pubs2' will be retried.

for 1 is not supported" may be reported on Solaris 10 setups unnecessarilly.

when JavaVM activities have left the PCI Bridge in a stale state previously. This core

dump file is not useful to Technical Support.

'memmove' may be reported in the error log while executing

a query requiring to access a partitioned table which partition

conditions had to be recreated from source text and statement

cache is enabled in the server.

dispatching. This is a fatal error, etc." when the PCI is unavailable with a stack trace

in the error log showing modules "jvminit", "SYB_CreateJavaVM" and "terminate_process"

although the stack trace information is unnecessary.

methods which perform lots of ASE table stored data manipulation or direct Client related data

processing. Although it is known that current execution performance is limited with respect to

what can be done it was found that certain COTS JavaVM instrumentation flags caused huge impact

on top of the current model. These instrumentation flags should not be enabled within customer

/ production environments.

In the case of a round robin policy any backup server should be configured in the interfaces file to run on any instance as supported by the documentation:

http://infocenter.sybase.com/help/topic/com.sybase.infocenter.dc00768.1503/html/ase_ce_ug/ase_ce_ug8.htm

Dedicated � each instance in the cluster is assigned a specific Backup Server.

Round robin � instances are not assigned a specific Backup Server, but when they perform a dump or load, the cluster assigns a Backup Server for the job according to availability.

with a stack trace containing function names alloc_dirtychains() and LeWkTbl::crtWkTbl().

on usage of the device immediately after creation of the device using 'disk init' for huge database devices.

with configuration option "enable literal autoparam"

active in some circumstances the execution of a query

containing a big literal binary value may be aborted

and the message "SQL Server system exception (0xc0000005)

generated by a storage access violation" together with

a stack trace which includes the modules 'memcpy' and

'mkconst_desc' may be reported in the error log.

during the upgrade process.

CgpPop::_CgpPePredBmToScalar visible in the stack trace in the error log.

Query involving outer join onto a virtual expression from a sub-query containing the column

involved in the virtual expression in a further outer-join clause.

A virtual expression is generally the result of a built-in function (isnull etc)

of multi databases transaction needs to rollback.

at the end of each statement because of which monProcessActivity had to be used for

tracking i/o activity for running query. The physical_io should report pbufread\pbufwrite

which are incremented as each i/o completes.

lock is always held.

function find_dup_corrname().

interpolation costing. This is a session-level counterpart of Traceflag 15310. It is OFF

by default for all optgoals.

CT-LIBRARY error:

ct_config(SET,IFILE): user api layer: external error: An illegal value of -9 given for parameter buflen.

when they are run with SET ANSINULL OFF could return different results

from Adaptive Server 12.5 when the results from Adaptive Server 12.5

are considered correct.

statement. The ASE errorlog will report a stacktrace involving module s_normalize. Also under some very rare circumstances on a heavily loaded system, there a risk of random memory corruptions.

errorlog while terminating ASE process.

with column based on user defined datatypes if the datatype definition is changed between

execution of the procedure.

LoadScore = ConnectionsWeight*ConnectionsLoad/100

+ CPUWeight*CPULoad/100

+ RunQueueWeight*RunQueueLoad/100

+ IOWeight*IOLoad/100

+ EngineWeight*EngineLoad/100

+ UserDefinedWeight*UserDefinedLoad/100

UNIONALL over DISTINCT query and the DISTINCT select list has two char type

columns which are equijoin'ed together. This would only happen when

the optcriteria "opportunistic_distinct_view" is off (by default it is on)

or the DISTINCT select list also has some expression (including subquery).

if an underlying table contains a column name longer than 31 characters.

This, in turn, may prevent certain data values from being eliminated before the comparison and cause an Arithmetic overflow. This CR is raised to see if improvements can be made in the allrow_mix to filter out "illegal" value.

memshrink() may occur after a 701 error.

while such an index already exists (for example, orphan index left over or index created first

with sp_reptostandby then again with sp_setreptable), the index re-created is corrupt.

The command 'dbcc tablealloc' reports error 2525.

Run-time errors 605 and 1133 may also be reported.

from that of version 12.5. In some cases, this may result in or-clause

predicates that reference only one table to not generate new

optimizable clauses.

Support has been added to revert the same table predicate transformation

behavior to be more similar to 12.5. This can enabled using trace commandline

flag 15333.

type column under non-binary sort order when merge_join is disabled. The EXISTS subquery's correlated predicate contains only equijoin.

Specifically, "PLAN FOR ONEROW" scans are not placed at the start of the

join orders. This only occurs when the query in run in "restricted compatibility"

mode as indicated by @@qpmode=2. See the "Using 'Compatibility Mode' In ASE

15.0.3 ESD#1 For ASE 15 Migration" technical white paper or the "Sybase� Professional Services - ASE 15 Migration Strategy" paper for more details on restricted

compatibility mode.

circumstances pertaining to a TSQL outer join with an or-clause that contains a subquery

that references an inner member of the outer join.

02:00:00000:00000:2009/10/14 15:44:04.07 kernel heartbeat message to server 3

failed, error status 4.

02:00:00000:00000:2009/10/14 15:44:09.10 kernel Instance 3 failed.

03:00:00000:00000:2009/10/14 15:44:04.07 kernel heartbeat message to server 3

in database 2 in DROP state." together with a stack trace which includes

the modules 'lwp_drop' and 'des__markdestroy' may be reported in the

error log when a session is trying to release a dynamic SQL statement

(previously prepared) and concurrently another session is trying to access

the same prepared statement, for example using the OBJECT_NAME() builtin

with the id used by the internal procedure created for the prepared

statement.

sets the parameters of a store procedure with a different datatype as defined in the store

procedure with NULL value. Command line traceflag 7741 could be used to allow the execution

of such application with success when 'statement cache' is configured.

to force group inserting for 12.5.x ordering compatibility (or using group_inserting

abstract plan operator, so not specific to trace 450), a query may hit error

257 (implicit conversion error) if group by clause (with having clause) contains

mixed datatypes which cannot be implicitly converted to one another.

for the same session.

has one or more BIT columns even though the option, allow_dup_row is not set. For duplicate

rows to be inserted, the total number of BIT columns in a row must not be an exact multiple

of the number of BITs that can be stored in a BYTE (8) - this is a necesary, but not

sufficient condition.

was not quiescent when the dump was performed ..." is reported when loading the dump of a database

that has a durability of either NO_RECOVERY or AT_SHUTDOWN on a platform that has a different byte

order to the platform on which the dump was taken. This is despite there being no activity in the

database at the time the database dump was taken.

generates a very wide worktable.

Solution is to fix the environment variable setup scripts like SYBASE.sh/SYBASE.csh not to include /usr/lib in the LD_LIBRARY_PATH (or /usr/lib/sparcv9 in LD_LIBRARY_PATH_64).

being compiled and Adaptive Server is running out of memory

for procedure cache the message "Stack overflow detected" may

be reported in the error log together with a stack trace which

includes the modules 'pars_release_resources' and 'recvhost'

and may result in Adaptive Server being shutdown automatically.

for example a query such as the following

select id from sysobjects c

where (id in (1,2) and cache = 10) OR (name="edwin" and uid = 5)

to be the dbid and dbname where the stored procedure executes.

CgpGenerateSQL::_findVariableValue.

the module responsible for physical I/O protection may not be correctly installed.

if any of the tables are roundrobin partitioned. The initial

plan generated by the alternative greedy algorithm may not be able to be

applied due to incompatible parallel degrees between a join's

input sets. This problem can be diagnosed with

show_abstract_plan tracing where "fail" will appear

prior to completing the alternative greedy algorithm phase.

requires an implicit conversion and the build input set has a column

with null values, then a 530 error may be generated during execution:

Msg 530, Level 16, State 3:

Line 2:

Attempt to insert NULL value into column 0 in work table (table id 4294967295); column does not allow NULLS. UPDATE fails.

Command has been aborted.

of the first OJ is a derived table or view may return wrong values.

This could occur only if the second OJ uses a hash or merge

join strategy. Here, first and second are referring to the relative

join order processing. Using the following query as an example:

select a.c1 from

a left join (select distinct b1 from b) as dt on a.a1 = dt.b1 -- OJ #1

left join c on dt.b1 = c.c1 and a.a1 = c.c1 -- OJ #2

Here, OJ #1's inner member is the derived table dt. OJ #2 uses

the result of OJ #1 in the on-clause "dt.b1 = c.c1" and these

columns are transitively associated to OJ #1's result. All of these

on-clause conditions must exist for the problem to occur. In that

case, incorrect values for a.c1 may be returned.

access based on a chain of ownership after the owner's alias is

removed from both databases and his login dropped.

as a variable rather than a literal, i.e. execute @procname. A workaround is to convert

this to dynamic sql, i.e. execute (@procname)

with a durability of NO_RECOVERY is being recovered, a 973 error

"The database '<template_dbname>' is currently unavailable. It is being used

to create a new database." may be reported in the error log.

Configuration Guide Adaptive Server� Enterprise 15.0.3 [ Windows ]

page 71 has the following ... ( the mention of gb108030 and big5hk is missing ) ...

...

Table 6-3 lists the simplified Chinese character set:

Table 6-3: Simplified Chinese character sets

Character set Unilib required Description

eucgb X EUC GB encoding = Simplified Chinese character sets

cp936 X Microsoft Simplified Chinese character sets

Table 6-4 lists the traditional Chinese character set:

Table 6-4: Traditional Chinese character set

Character set Unilib required Description

cp950 X PC (Microsoft) Traditional Chinese

euccns X EUC CNS encoding = Traditional with extensions

big5 X Big 5 Traditional Chinese

...

page 71 should have the following ... ( gb108030 and big5hk were added ) ...

...

Table 6-3 lists the simplified Chinese character set:

Table 6-3: Simplified Chinese character sets

Character set Unilib required Description

eucgb X EUC GB encoding = Simplified Chinese character sets

cp936 X Microsoft Simplified Chinese character sets

gb18030 X PRC 18030 standard

Table 6-4 lists the traditional Chinese character set:

Table 6-4: Traditional Chinese character set

Character set Unilib required Description

cp950 X PC (Microsoft) Traditional Chinese

euccns X EUC CNS encoding = Traditional with extensions

big5 X Big 5 Traditional Chinese

big5hk X Big 5 with HKSCS extensions

...

indicating that a page being recovered has a timestamp older

than what we expect at that point.

DBISQL returns a syntax error.

when using trace flag 450

If the session-level 'set option show_missing_stats on' is used it is also possible to hit this bug, stack trace will include VbmManager::_VbmPutBm() and GttTable::GtGatherNoStats()

which is 1962 bytes." is raised when installing the system stored procedures

sp_help, sp_helpdb, sp_helpindex and sp_help_rep_agent.

and outer join, may get merged into an OR predicate

which has other subqueries, can result in an incorrect answer.

Trace flag 16902 will turn off this change since there may

be some performance degradation.

READ ONLY. " may be raised when trying to access a partitioned table in an archive database.

module leading sometimes to the syntax errors on the replicated site.

with esd version #2 or lower, error 2736 occurs if the user is not a

named user in the loaded database and the database has a guest account.

and the spid terminated when fetching a cursor after an update at isolation level 0.

The cursor would need to be sensitive (likely to involve a hashing operator)

IN item list is present in different clauses of the query.

be respected if the query contains nested sub-queries and two or more tables are joined

along with one table also being involved in a left join to another reference of the same

table.

of LWP with distinct and outer joins.

when ASE incorrectly tries to flush index pages to disk in an archive database,

without having remapped them to the modified pages section.

terminated under rare circumstances involving a reformatted correlated subquery with group

by\having clause which references a column in the outer query. Other entries in the stack

would include CgpGroupHashing & CgpStoreInd

which do not return a result set.

an error 156 during attempted execution of invalid 'create procedure' statement contained

within begin..end block.

tab.col != constant1 and tab.col != constant2 ... then there is a selectivity

estimation error proportional to the number of NULL values in the column. Also,

for the specific case of tab.col != NULL, the estimate w.r.t. the ANSINULL

state was inverted, i.e. 0.0 if ANSINULL was off, and the number of NULL

values if ANSINULL was on.

They may progress in parallel on a database; may result in corruption.

to 'DEFAULT' after reboot.

add the same login and suid again, the attempt fails with error:

"Fatal protocol error. xact_beginupdate() API was invoked in the wrong context..."

or parameters. Stack trace is produced indicating module opt_like_and.

into a database on Adaptive Server 15.5 (non-CE). In order to ensure disaster recovery then

either load the dumps into a database on Adaptive Server 15.5 CE or use Replication Server

or sybmigrate, the Sybase Migration tool.

for DISTINCT view self join. Due to the self join,

DISTINCT view has to be materialized using a Store operator.

The DISTINCT view select list has a char type

column which is equijoin'ed with another char type column.

This would only happen when the optcriteria "opportunistic_distinct_view"

is off (by default it is on)

or the DISTINCT select list also has some expression.

subsequently hanging other tasks in ASE. The unmirroring task cannot be killed and the ASE server

needs to be restarted.

can result in server shutting down.

when the join is pushed down into the materialization step.

The NULL = NULL join is created when an explicit predicate such as

r.a = s.a OR (r.a IS NULL and s.a IS NULL).

The NULL = NULL join can be implicitly created in an extended column

aggregate (a non ANSI sybase feature) such as

select r.b, avg(r.c) from r group by r.a

The NULL = NULL join semantics can be turned off, and instead

normal join semantics could be used instead, resulting in a

wrong answer.

Arithmetic overflow during implicit conversion of BIGDATETIME value '

Invalid pointer param number 3, pointer value 0x9cf500d408f1

This CR is to change the return value to be either current_date() or raise an error stating 'not supported' when a pre-15.5 client is detected.

adding customized Property tab page (e.g. for Element in XSM).

With the new tab page, some buttons do not work as expected, producing empty windows or

no response at all. E.g. it is discovered that those buttons of "reference object" and

"type object" on it are fail to browse and only get an empty list. While the default

General tab has no problem.

language ) "xsd.xsl" by adding customized Property tab page (e.g. for Element in XSM).

With the new tab page, some buttons do not work as expected, producing empty windows or

no response at all. With the customized property tab page, there seems no way to add Data Source and data mapping with the source.

may raise a 530 error (NULL can't be inserted in worktable)

if the query uses a distinct_hashing strategy and the non-correlated

subquery returns a NULL.

position 2 of the extrainfo section of the audit record if available. Otherwise the

Grantee and Command Type will be written.

in allrows oltp compare to compatibility mode.

This needs to be stated in the docs.

installations. The following error is generated:

optdiag simulate statistics qts_db -Usa -P -T1

OptDiag/15.5/EBF XXXXX SMP/B/x86_64/Enterprise Linux/asearies/2431/64-bit/OPT/Mon Jan 11 00:38:22 2010

...

Server Message: - Msg 207, Level 16, State 4:

Invalid column name 'instanceid'.

Internal error: ct_results() returned result type CS_CMD_FAIL.

Optdiag failed. Check for error messages above.

in-order rows if a general OR optimization is used or if the non-covered

index scan is not clustered. The in-order row estimate

is used to help reduce physical IO estimates when good cache hits

occur.

fetched from a cursor that involves the same table again. This situation is only likely to occur when

index split happens due to the insertion of data row.

password_random will cause a SIGSEGV resulting in a stack trace in ASE 15.5

if the /dev/random, /dev/urandom, and /dev/srandom do not exist.

a table from a DISTINCT view or derived table on some outerjoin inner table column,

provided there is an extra WHERE clause on the outerjoin inner table column before

the INSERT. The extra rows are due to not including the WHERE clause in the final plan.

during implicit conversion of DATE value to a DATETIME field) on a row containing NULL in a nullable DATE column.

configuration failed because "The sum, (number of user connections) + (number of devices *

max online engines) + (number of remote sites) + (max number network listeners) +

(max cis remote connections) + (number of java sockets), must be no greater than '<value>'", where the 'value' at the end is not correct.

in the PCA/JVM, through either the execution of stored procedure sp_jreconfig "disable",

"pca_jvm_netio" or manually does not prevent RMI socket operations from working despite

the property being set to disabled. Consequently a Java RMI client will continue to be able

to connect to a remote RMI server to execute some functionality over the network, although it

should not. To avoid this problem remove the unused RMI class(es) from the database.

The pca_jvm_netio argument is enabled by default.

This error does not occur if the server is started as a windows service.

1. server's default sortorder is non-binary single-byte sortorder

2. like pattern is a constant literal

3. like pattern string length is the same as column data length

4. column is either varchar or char

's_begupdate' together with a stack trace which includes the

modules 's_execute' and 'execproc' may be reported in the

error log when a stored procedure executes a DML statement

which was compiled at runtime and affects a view which had

defined an Instead of Trigger.

1. Unichar

2. Univarchar

3. Bigtime

4. Bigdatetime

followed by Multibuffering subprocesses died

seen is as follows:

Expected to find the descriptor for object %d in database %d in CLONED state.

log__newpage: Page <pageno> is in DESTROY state prior to bufawrite.

plc__flush: Page <page> is in DESTROY state prior to bufpredirty.

This should be fixed or documented in the install guide

The spreading of dbt_dbname field corruption from one instance

to other instance is prevented. The DBCC DBTABLE command is

enhanced to set a new database name in the DBTABLE on all cluster

instances.

... ks_falloc: Cannot allocate using ubfalloc(rnetmempool, 8192)

Optimizer using Table scan because cost is cheaper than OR Strategy

surrounding lwp (statement cache) and procedure execution.

Currently the figures (which will generally be reported via sp_sysmon) can be a little

misleading under certain circumstances.

an outerjoin with a WHERE constant equality clause (col=<cnt>) on an outerjoin

inner table column and DistinctSorting is in the final plan.

This will only effect non-binary sort orders.

"CgEvals::_CgCompVcol" may be reported in the error log followed by a stack

trace during the execution of a query involving UDF/ADT constructions.

The modules "jvmfini" and "BRIDGEjvm_AbortJavaVM" will also be reported

resulting in the current session terminating and the JVM aborting. Alternatively

this may cause ASE shutting down itself.

It does not stacktrace on dump and load migration of sybsystemdb onlinedb.

to execute an external web service through the right-click->'Execute Web Services dialogue

box on the 'Web Methods' tab under the 'Web Services' node.

with "ct_connect() failed", "jsd_MakeConnection() failed", "jsj_AddConxs() failed",

"jsj_AddConxs() failed" and "Initialization failed initializing connection pool" messages

in jsagent log file.

with a stack trace which includes the modules 'clean_process' and 'kill_proc'

may be reported in the error log when a session executing a query involving

session temporary tables was killed by another session and statement cache is

enabled in Adaptive Server. Under some circumstamces a 6103 error, "Unable to do

cleanup for the killed process; ..." may also be reported after that.

selecting an aggregate function such as count() from a DISTINCT derived table based on an

equijoin between local and proxy tables and the equijoin is on char/varchar type columns.

(<module name>) Diagnostics for error 707 state 10: <diagnostic info>

is reported in the error log together with a stack trace.

An additional single line of diagnostics is reported prior to the message

and stack trace above, when the 723 trace flag is turned on.

may lead to subsequent error 1601 and error 8601on the same instance, and bunch of user tasks can't

proceed on one or multiple instances of the cluster when the situation occurs.

a stored procedure if the procedural statement is compiled in the deferred mode.

involved on some of the columns or expressions are used for which histograms do not exist.

even if the master device is the file system and the platform supports direct i/o.

then it will not be recognized as a dense frequency cell, which means that

for this specific date, the range density would be used for an estimation

rather than the weight of the frequency cell in the histogram.

fix is enabled under optcriteria cr619965 which has been associated

with optlevel ase1503esd4 for all optgoals

that 2792 is a trace flag used to make the concurrent tasks sleep when reorg rebuild/AMC is under way instead of throwing 8233 error.

normally performed when an ASE task disconnects from the COTS JavaVM,

stale entries for UDF/ADT objects may be left behind and could

be processed mistakenly when the COTS JavaVM migrates to another

engine resulting in unpredictable JavaVM processing and sometimes

in ASE shutting itself implicitly.

MFC 15. They are mapped to ASE DATETIME datatype instead.

A workaround fix is provided to enable the mapping and traceflag 11238 is

used to turn on the fix.

pool and can not be killed.

The correct formula for the precision should be 憁ax(s1 + p2 + 1, 6) + p1 - s1 + s2� and for scale 憁ax( s1 + p2 + 1,6)�

cpuinfo , lmutil and sysam.

1. LANG=ja_JP.utf8

2. setup -console option used.

a stack trace which includes the modules 'closetable' and 'pg__allocate' may be

reported in the error log when a DML command has been previously run and hit

error 1105.

switch 3635 turned on.

an administrative function and / or sproc to be able to display the info returned by ...

show_role()

sp_activeroles ... this gives both show_role() and user_defined roles

...for spids other than the current administrator spid.

In other words ... let's say there are 3 active users in the server...spids 12, 17, and 21 The administrator is logged in with isql and has spid ... 17. OK right now this administrator is only able to get show_role() to return the active roles for spid 17. What the customer wants is to show the active roles for spids 12 and 21 also.

with the chained transaction mode may fail with a timeslice error being

reported in the errorlog showing the module xls_test_set_lrdata and

dbts_commits_addxdes.

"ct_connect() failed", "jsd_MakeConnection() failed", "jsj_AddConxs() failed", "jsj_AddConxs() failed"

and "Initialization failed initializing connection pool" messages in jsagent log file.

is part of an EXISTS join (e.g. under IN/Any subquery),

signal 11 stacktrace could happen in CgpPop::_CgpPePredBmToScalar() for

distinct operator (CgpDistinctHashing/CgpDistinctSorting/CgpDistinctSorted)

when the following three conditions are all true:

1. The DISTINCT view/derived table has at least one expression SELECT column;

2. The parent SELECT DISTINCT columns are only a subset of the DISTINCT view columns.

Or if the EXISTS join is on one of the DISTINCT view columns;

3. One of DISTINCT view columns which is not part of 2) has a predicate on it

in the parent query block.

during attempt to free memory at address 0x<address>. Please consult the

ASE error log for more details." may be reported for an ASE that uses

a procedure cache whose size combined with the size of the statement cache

exceed the value of 4G-bytes.

of the transaction but before post commit work started. This CR is a feature request to make this default behavior and retire the traceflag.

normal mode) when it references a view and there is another view of the same name

under the user executing the procedure.

With 'set showplan on', you will see:-

QUERY PLAN IS RECOMPILED DUE TO DEFERRED_COMPILE_FAIL, TABMISSING.

THE RECOMPILED QUERY PLAN IS:

in circumstances where the procedure has been recompiled due to a deferred compilation

failure.

For example:-

1> exec p1

2> go

(1 row affected)

c1

-----------

(0 rows affected)

(return status = 0)

1>

Instead of:-

1> exec p1

2> go

c1

-----------

(0 rows affected)

(return status = 0)

1>

in permission error 10351 if the caller of the stored procedure is not the owner of the

udf.

derived table, the expanded derived table columns may be placed within the derived table

definition as oppose to the outer query.

This could lead to an error 107 when re-creating the procedure from the text stored in

syscomments

an outerjoin between a GROUP BY view/derived table and other table

when there is a constant sarg in the view for the equality joining

column. Null-supplied rows from the outerjoin may be incorrectly

eliminated.

with unique index. Trace flag 16842 can be used to track the SQL that caused the

2601 error.

is reported in errorlog and login failures prevent additional user connections.

There is no guarantee to evaluate and execute predicates in left-to-right order.

For example, if you run a query having "WHERE x != 0 AND y = 10 OR z = 100", Adaptive Server does not guarantee to evaluate and execute "x != 0" first.

Adaptive Server can evaluate and execute predicates in any order.

{BINARY/VARBINARY} or a key pairing of {BINARY(length M)/BINARY(length N)}

and the binary values differ only by the number of trailing 0s, then these

rows won't be returned.

while accessing ... object 'syslogs' (8) ..." together with a stack trace which

includes the modules 'rep_agent_thread' and 'xsc__syslogs_getnext' and 'wrongpage'

may be reported in the error log by the Adaptive Server Replication Agent task.

reported together with a stack trace which include the module 'writetext' when executing

the command 'WRITETEXT WITH LOG' on a replicated text column which does not have an index for replication.

in database '<value>'. All rows that qualified during the first phase were not visited

during second phase of dml." may be reported in the error log together with a stack trace

which includes the modules 'dodeferred' and 'dol_def_update' if an UPDATE command has

been previously run against a data-only-locked table.

The message needs enhancement to prompt user to verify Backup Server version compatibility with ASE.

Backup Server error message:

Backup Server: 3.31.2.3: Improper parameter count 6 for RPC bs_begin_load;

expected a minimum of 3 parameters and a maximum of '5' parameters.

The ASE stack trace gives the impression that the problem is complicated. The simple solution is to use the newer version of the Backup Server.

policy, the instance level physical lock on a page may be released

without ensuring all local users of the page have finished working on it.

Subsequently, the Adaptive Server Cluster Edition will report the

following 892 error:

"Process <spid> is holding a latch on buffer <buffer-address>

(dbid: <dbid>, pageno: <pageno>) without holding a physical

lock on the buffer".

select * from t1 where t1.a1 = (select max(b2) from t2 where b2 <= 10 and a1=b1)

should behave the same w.r.t. subquery index selection on an index on t2(b1 asc, b2 desc) as

select * from t1 where t1.a1 = (select max(b2) from t2 where b2 <= 10 and 5 =b1)

The definition of the @server parameter of ASE15 to sp_passthru procedure was extended to

varchar (255). And the errmsg parameter was also extended to varchar (1024).

phases of the operation.

In Example 11, the object name has been ommited from the command syntax. The first parameter after the stored procedure name has been omitted. This should read:

Example 11

Switches the deallocation for text and image space on using dealloc_first_txtpg:

sp_chgattribute "mytable","dealloc_first_txtpg", 1

To switch the feature off:

sp_chgattribute "mytable","dealloc_first_txtpg", 0

one of the other instances crashes.

or 1 into systabstats system table if the ASE page size is more than 2K.

reporting large I/O prefetch being used when eventually it was not

sqlsrv.avail_physical_memory: PUT_THE_AVAILABLE_PHYSICAL_MEMORY_FOR_ASE_IN_OPTIMIZATION

sqlsrv.avail_cpu_num: PUT_THE_AVAILABLE_NUMBER_CPU_FOR_ASE_IN_OPTIMIZATION.

actively inserting and modifying data.

Many errors will be produced with -6989 in the error message.

the size of single page of 16384 bytes" or some random stack

traces may be reported in the error log when a stored procedure

(or SQL user defined function) which contains an insensitive

cursor involving an expression which results in a size bigger

that 255 bytes is executed by reusing a plan generated by a

previous session.

will not exclude the entry in sysindexes for the table itself.

This could lead to a failed drop index on running the script in scenarios whereby a table

has the same name as one of its indices.

and the ASE is configures with CIS services to access non ASE servers.

under a scenario whereby a rollback has been performed on an insert via an insert trigger

and a further insert..select then references a text column from the inserted logical table.

It is not supported to reference text (or other LOB off-row storage) datatype columns in

the inserted\logical tables used by a trigger.

inserttext() will be seen in the stack trace.

From 15.0.3 ESD#4, we will raise error 2116 and prevent the trigger from being created.

then there is a possibility of a wrong answer when a DISTINCT, ORDER BY, or GROUP BY is

used in the query with = NULL .... or an equi-sarg like = @v1.

The behavior is as if ansinull is on, which would eliminate NULLs.

the following syntax is incorrect ...

sp_lmconfig

[ 'edition' [, edition_type ]]

[ , 'license type' [ , license_type_name ]]

[ , 'smtp host' [ , smtp_host_name ]]

[ , 'smtp port' [ , smtp_port_number ]]

[ , 'email sender' [ , sender_email_address ]]

[ , 'email recipients' [ , email_recipients ]]]

[ , 'email severity' [ , email_severity ]]

the following syntax is correct ...

sp_lmconfig

[

[ 'edition' [, edition_type ]] |

[ 'license type' [, license_type_name ]] |

[ 'smtp host' [, smtp_host_name ]] |

[ 'smtp port' [, smtp_port_number ]] |

[ 'email sender' [, sender_email_address ]] |

[ 'email recipients' [, email_recipients ]]] |

[ 'email severity' [, email_severity ]]

]

CR 623307 fixed a problem in which a 692 error might be reported by

the Adaptive Server Replication Agent task. The current fix besides addressing

that performance issue it ensures that a scanner of the log is able

to recognize a stale log page and refresh it.

to a proxy column is executed repeatedly.

Install Linux kernel patch for this bug to resolve the memory leak.

This makes it very hard to find differences

"xsd.xsl" by adding customized Property tab page (e.g. for Element in XSM).

With the new tab page, it includes the original NAME property in Element entity.

It is discovered that modifying the NAME property, the corresponding CODE property will not be changed together as the default General tab page.

especially for partitioned tables, when upgrading the database from any release to

15.0 or higher.

sides of the UNION ALL may generate a SEGV during the cloning of the hash

operator's input aligned data row. This will occur if there is

a join between projected columns of differing lengths in at least one

of the UNION ALLs select statements.

may happen when executing a procedure with a cursor with union operator and if the query returns

constant literal values.

error 325, Level 18, State 1.

or use sp_help ..." with Adaptive Server 15.0.3 when a SELECT command is run against

an archive database user table and the archive database dump was taken with Adaptive Server

version prior to 15.0.

This is due to the current Axis 1.3 implementation within ASE Web Services.

You may receive:-

Msg 15200, Level 20, State 0:

Server 'ws', Line 1:

No web methods mapped to proxy tables for the WSDL URI

on running sp_webservices 'add' against the wsdl

Since 11.9.2, to facilitate concurrency we raise a 2601 error on insert into a unique index

on a datarows locking table if that key already exists (whether committed or uncommitted).

There is no risk to data integrity.

The request is really to document this from the perspective of the relaxation of the

isolation level for this scenario.

optcriteria cr628311 associated with optlevel ase1503esd4 or trace flag 16909 will turn on the change to correct this issue.

storage access violation on Windows) in memset() and stacktrace. This is due to

fragmentation in the Pss Session Heap memory pool

particular SET ANSINULL setting (ON or OFF) and later a session

with different SET ANSINULL setting executes the same stored

procedure then the stored procedure is recompiled. This could

cause excessive recompilations if the same stored procedure is

executed concurrently by many sessions with different ANSINULL

setting.

After this fix, boot time trace flag 753 is obsolete and large allocation auto-tune

in memory manager is turned off by default. Use boot time trace flag 758 to turn on

when necessary.

from ase1503esd2), and your server is configured for parallel, you may see a degradation

in performance for certain ORDER BY queries due to the avoid_bmo_sorts optimizer criterion being

enabled. optcriteria cr628845 or trace flag 16912 will correct this issue.

Unable to allocate work table for processing group by or distinct clause during query optimization. Maximum limit (46) of work tables is exceeded. Simplify the query and retry.

the workaround is relatively complicated ...

logical page <page> belongs to object with id ..."

may be falsely reported during INSERT. The modules

pg_allocate() and pg_latchoam() will be reported in

the stack.

on for the CIS connection wish mean double quotes is not allowed.

Environment:

Sybase ETL 4.9

Windows 2003 Server

Problem Description:

Row delimiter of generated Load Script is incorrect.

Description of the problem:

Customer tried to use IQ Loader File via Load Table component, and they set 2 properties like following.

- Line Delimiter of Text Source Definition : CRLF

- Row Delimiter of Properties : CRLF

Next, run "Generate Load Script" from IQ Loader File via Load Table comtext menu.

And Pushed Load Script Button.

Then generated load script is....

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

LOAD TABLE DBA.test3

(

[col1] ',',

[col2] 0x0d

,FILLER(1)

)

FROM 'D:/work/crlf_test2.txt'

QUOTES OFF

ESCAPES OFF

BYTE ORDER NATIVE

BLOCK SIZE 10000

WORD SKIP 0

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

"0x0d / FILLER(1)" was generated insted of CRLF (0x0d0a).

If "0x0d" is included in source data, above load script doesn't work correctly.

f for IQ performance, I think that using "row delimited by" load table command gets more better performance, like following.

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

LOAD TABLE DBA.test3

(

[col1] ',',

[col2] 0x0d

)

FROM 'D:/work/crlf_test2.txt'

Row delimited by '\n' <-----

QUOTES OFF

ESCAPES OFF

BYTE ORDER NATIVE

BLOCK SIZE 10000

WORD SKIP 0

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

Because if use "row delimited by" statement, IQ tries to load table as multithread.

Results of my test, using "row delimited by" situation is very good performance than original generated load script.

RECOVERABLE) dropped during query optimization." may be reported

when running a SELECT INTO from a view that had to be internally

recreated from its source text, for example, because the database

was upgraded from a previous version.

Server '<SERVER>', Procedure 'sp_sysmon_netio', Line 260:

Arithmetic overflow occurred.

May be seen when running sp_sysmon.

It is caused by a sum aggregation on an int column (that is required to be int to be in

line with system tables) exceeding max int.

crash the server with signal 10 (SIGBUS).

disk cluster such that two or more surviving nodes are left and failover recovery was

needed for a multi-database transaction that was found in a prepared state. If the

database is replicated, it is possible that the stale log pages will result in errors

during replication.

involving OR predicates (i.e. IN lists) with equi-sargs on the same column, and joins to columns

on differing datatypes

e.g. r.intcol in (6,10, ...) AND r.intcol = s.numericcol

This should only happen if constants are used without literal parameterization,

i.e. using variables in the OR predicate should avoid this error.

0x<value>" together with a stack trace which includes the modules 'rep_agent_thread',

'ra_process_lr', 'ra_trace', 'scerrlog' maybe be reported in the errorlog when RepAgent

thread tracing is active.

specification of separate log\data devices for tempdb.

Currently we only have these parameters available:-

sqlsrv.tempdb_device_physical_name

sqlsrv.tempdb_device_size

temporary directory used by InstallShield (equivalent to setting -is:tempdir on the

command line for the installer).

$SYBASE/ASE-15_0/init/logs>

find . -name "*" | xargs grep depre

...

./sqlupgrade0601.001: Stored Procedure '' uses deprecated builtin functions as:

...

"PCI(T1): -> PCA_JVM: JVMLOG 'FATAL ERROR in native method: Bad global or local ref

passed to JNI'" followed by "PCI(T1): JVM_ABORT called - ignore", OR

"PCI(T1): DEBUG: pci_command_stop; Request to stop slot 1" may be reported

in the error log during the execution of a query involving UDF/ADT constructions.

This error is caused because the COTS JavaVM integrated PCA/JavaVM cache gets corrupted.

This will result in either the current session terminating or ASE shutting down itself.

In order for the current fix to be effective and prior to start the COTS JavaVM,

the PCA/JavaVM cache needs to be disabled by the SA user by running in the sybpcidb

database the command: sp_jreconfig "disable", "pca_jvm_sigcache_enabled"

the scalaragg is subsequenctly used in a group-by clause and the

result of the scalaragg grouping column is is subsequently

joined to itself, then a SEGV could occur in _CgReplaceVars().

Note extended column semantics will create an implicit join between the

grouping instance of the scalaragg and another instance of the scalaragg.

The following extended column query is an example of the problem:

select s.c3

from (select count(r.c1) cntem, r.c2, r.c3 from r) s

group by s.cntem, s.c2

It is believed to be fixed in kernel version 2.6.5-7.322

Without the fix:-

kernel BUG at fs/aio.c:733!

may be seen in the messages file during an async i/o operation.

(Object not found) if the procedure was created by the user with 'sa' authorization

under impersonated user with setuser 'username' command and the execution is done by the

real connection with 'username'.

'Creating a Response File' actually points to the PC Client 'Creating a Response File' page

not the server page, as such it contains references to setupConsole as oppose to setup.bin

We may see routines endscan() and trigger__target_confirm() on the stack trace.

by multiple sessions, ASE may run into a situation where the signature cache of the

PCI integrated JavaVM gets corrupted causing the report of inconsistent error messages.

In some cases this may result in the JVM aborting or ASE shutting down itself.

during an online database which contains sysroles.id equal to -1

(due to a prior load at some point that didn't have the server roles in syssrvroles.)

e.g. if you try to distinguish what columns are involved in the index keys from the

index list in Sybase Central then when there are multiple columns involved in the

index, they are all concatinated into one long meaningless name ... ??? ...

for example there is the index ... nc2syspartitions on ... master.dbo.syspartions

with index_keys of ... id, indid, name -- but in Sybase Central this information is displayed as idindidname ... ??? ...

This may be followed by SIGSEGV, SIGBUS or timeslice error and lead to ASE hang.

When this method is used and one of the queries that has been included in the resource file returns no rows,

Migration tool will report an error and the migration process stops.

character semantics for comparisons of table names, correlation names

index names, work table names, disk names, view names, node names.

This could result in name mismatches with mixed case for nocase installations.

has the outer position in the join.

to which a named cache is bound implicitly or explicitly.

INT instead of STRING in the config file, which may cause server boot failure.

These inconsistencies should not be reported and may disappear on the second attempt of dbcc checkstorage.

columns in the select list of a GROUP BY aggregate

which are not in the GROUP BY list) which in turn

reference union views can lead to a signal 11 in

markColForStat or ind_get_ides.

a dbisql dialogue box stating:-

"The runtime environment could not be found in ."

This is due to a call made by the installer to terminate the dbisql fast-launcher.

The error message is raised as it runs the 15.0.2 DBISQL (which is version

ddlgen -Usa -Sservername -Ppw -Ddbname -TDB

DDLGen generates the following info for system view: sysquerymetrics

"-- Cannot Generate DDL for system view(s)."

00:00000:00013:2010/06/25 16:31:42.39 server Started Rep Agent on database, 'prim' (dbid = 4).

Unable to retrieve localized message for ID 1

This message is reported even nor SSL nor LDAP is used.

Even this message is reported, RepAgent thread connection to RepServer succeed and replication works as expected.

This message is raised by CSI layer.

CSI layer is not able to find its localized file to report a warning.

The message that should be report is the following:

00:00000:00011:2010/06/25 16:28:19.20 server Started Rep Agent on database, 'prim' (dbid = 4).

Sybase CSI Error Message: OpenSSL provider instances cannot be loaded concurrently with multiple, distinct, system policies due to limitations within OpenSSL.

CSI layer can't find its localized files because localized files are installed in directory "$SYBASE/locales/unicode/csi/en_us"

instead of "$SYBASE/locales/unicode/csi/en_US" (note the "US" in upper case)

concurrently

after to modify the 'plldegree', then the attribute change may be lost the next time the

housekeeper flushes the in-memory statistics.

inner table of an outer join and that expression is in the WHERE clause, then the outer join

can degenerate into an inner join (as seen in showplan) resulting in

an incorrect answer with null-supplied rows

being eliminated.

in the ASE errorlog for reorg rebuild <tablename>.

These should cover:-

reorg rebuild <table> <index>

&

reorg rebuild <table> <index> partition <index_partitionname>

Which would generate:-

Beginning REORG REBUILD of index '%.*s' on table '%.*s'.

REORG REBUILD of index '%.*s' on table '%.*s' completed.

&

Beginning REORG REBUILD of partition '%.*s' of index '%.*s' on table '%.*s'.

REORG REBUILD of partition '%.*s' of index '%.*s' on table '%.*s' completed.

in methods being executed unexpectedly due to class-loader structures getting

corrupted. Alternatively Java Null Pointer Exceptions may be raised within classes

unexpectedly. When global references get destroyed repeatedly, this will result

in the COTS JavaVM aborting.

ddlgen -Usa -S... -P... -D... -TD -TI -TLK -TKC -TR -TRI -TSGM -TTR -TU -TV -N% ,

ddlgen does not print a syntax error.

in the module 'rtrim_unichar' or 'lower_unichar' maybe be reported

in the error log when it is executed a query joining two or more

tables and using the functions RTRIM() or LOWER() on a column of

datatype UNICHAR or UNIVARCHAR.

even if you connect to the ASE with us_english language.

shutdown reporting the error 9202, Severity 20, State 0,

if the replicated table has a trigger and the SQL DML

within the trigger attempts to use SQL Statement

replication as well.

a show_cached_text() for the LWPs.

UTF8) could lead to a failed jdbc connection to the same case-insensitive server.

This will be seen at the client-side:-

java.sql.SQLException: JZ00L: Login failed. Examine the SQLWarnings chained to this exception for the reason(s).

SQLState: JZ0IB

Message: JZ0IB: The server's default charset of UTF8 does not map to an encoding that is available in the client Java environment. Because jConnect will not be able to do client-side conversion, the connection is unusable and is being closed. Try using a later Java version, or try including your Ja

va installation's i18n.jar or charsets.jar file in the classpath.

Vendor: 0

SQLState: JZ006

Message: JZ006: Caught IOException: java.io.CharConversionException: java.io.UnsupportedEncodingException: UnsupportedCharset

Vendor: 0

this is a request to print a error message.

<X> bytes is larger than the maximum size (<Y> bytes) allowed for this table"

may be seen during 'optdiag binary statistics' output when running under compatibility_mode.

is aborted due to a 1205 error, "Your server command (family id ...,

process id ...) encountered a deadlock situation. Please re-run

your command." and the procedure was executed by another stored

procedure that had a cursor opened inside a user transaction and

the option SET CLOSE ON END TRAN was ON then the next execution

of the same stored procedure that reuses the plan which was

aborted could behave ignoring the WHERE clause for the statement

that was aborted by the 1205 error.

was invoked in the current session or adhoc rows were inserted in master.dbo.sysattributes with object_type = "PS".

This can occur at server startup or at user authentication when attempting to

obtain the lock which is already held.

together with a stack trace which includes the modules 'lt_sqlrep_comp_policy',

's_compile_stmt' and 's_recompile' when a DML command is ran and the involved

queries are eligible for SQL Statement Replication.

Once it has been run with the abstract plan from 12.5.4, it completes, but as soon as you update statistics again, it is broken.

" Job Scheduler Agent entry for host '<host_name>' not found. " when the machine name(host_name)

for JSAGENT entry in the interface file is not same as name returned by `uname -a`.

unmounted and mounted a database that is using this device.

no open transaction.

by multiple sessions, ASE may run into a situation where the PCI integrated JavaVM aborts

after having reported "OutOfMemory" error messages. In some cases this may result in ASE

shutting down itself.

starting an ASE, sp_listener "start","ssltcp:<host>:<port>:'CN=<common name>'","<engine>"

will fail, with common name mismatch error.

expected blocking behaviour on a row that is inserted by an uncommitted transaction

if the inserted row is replacing a row deleted by the same transaction.

related to sysobjects with statistics on ASE 15.5

and the spid terminated when attempting to insert <table> (<column-list>)..select <col> from <table2> union/union all <select> in a procedure where <table> is created inside the procedure and <table> contains non-nullable columns that are not in <column-list>.

CreationDate\LastExecution\NextExecution dates to display a 'null' when viewing

'Scheduled Jobs' (requires Job Scheduler) in Sybase Central 4.3 through to 6.0.

The stack trace may contain oh_get_obfuscated(), copy_sqltext() and sqlpars().

dbccdb log fills and when cancel out of dbcc checkstorage may cause sigsegv and stackoverflow.

with configuration options "enable literal autoparam"

and "statement cache" active in some circumstances the

execution of a query containing a big literal binary

value may be aborted and the message "SQL Server system

exception (0xc0000005) generated by a storage access

violation" together with a stack trace which includes

the modules 'memcpy' and 'mkconst_desc' may be reported

in the error log.

the deferred parallel optimization mechanism. The default value is 0 (disabled);

a value of 1 enables this method.

Deferred Parallel Optimization is an alternative method for costing and selecting

parallel query plans. When enabled, the best serial plan is chosen using the

serial optimizer; this plan is then converted to a parallel plan, normally

maintaining the same join order and physical operators as the serial plan.

This method provides a parallel plan using less optimization time, while reducing

the risk that a parallel plan will be chosen which is inferior to the best serial

plan. With the full (default) parallel optimization mechanism, an inferior plan

may be chosen due to optimizer time out, or due to limitations of parallel costing.

1) update with setting constant data

2) update with setting data from other columns

3) update with setting the column itself

However, "by Log Marker" is another line item that has no description in the docs anywhere.

this values "NULL, 0 ,1"

because the target page is not a data page... pageno = 0 ..." may be reported during the

execution of LOAD TRANSACTION command using the WITH UNTIL_TIME option from a transcation

dump that was dumped using the WITH NO_TRUNCATE option.

When both referential_constraint_create_list_from_sql and table_create_list_from_sql are both set, both provide a list of objects to migrate. The list of foreign keys does NOT depend on the objects identified for migration by table_create_list_from_sql.

pagesize and the hash join resorts to a nested loops strategy, then

wrong results could be returned. The input row's lengths can be determined

by running with trace commandline flag 9529.

enough, user may hit error 8211 when try to access the table from remote node.

Msg 2056. Line 1, character 8290: Incorrect syntax with ' '" may be reported

in the error log as a result of uninitialized database id.

The conditions leading to this are data specific and rare.

This may only occur on a non-binary sort order server.

its own table when both tables are using same name and a previous access to the dbo

table was done.

may occur with a stack trace with the following message:

"ks_falloc: Cannot allocate using ubfalloc"

can't be referenced in other parts of the query. This is only to avoid syntax error.

different sortorder from the local column may get a wrong result.

an aggregate (or possible other virtual expression) and a further outer join may result

in a sub-optimal query plan.

SELECT list if the subquery is selecting from a single table/view/derived table

with a single column. It will function as SELECTing the single column explicitly.

stays the same. Each of the multiple items will be treated the same way as with a single SELECT item for the EXISTS subquery.

Backup API library.

be raised with sig11 from s_result_const_memfree->free__linkedcst->memfreeheap.

'ptn_pdes_install' together with a stack trace which includes the modules 'dorecover'

and 'db__create_sysaltusages_map' may be reported in the error log when an archive

database is being recovered.

This table should provide similar information to what is contained in monSysStatement

and monProcessStatement. The data that is captured at the statement level in

monSysStatement will be aggregated to the stored procedure level.

dataserver if a load of sybsecurity database had resulted in less tables than configured

for. The mismatch will be seen when executing sp_configure 'current audit table'.

update statistics command with the index name may raise a 2727 error.

'getvalue' together with a stack trace which includes the modules

's_retparams' and 'execproc' may be reported in the error log when

an INSERT statement is run on a table having an encrypted CHAR/VARCHAR

column, the configuration parameter "statement cache" is active and

the INSERT statement was previously cached but it had to be recompiled

or renormalized because, for example, all the query plans were in

use or the schema for the underlying table changed.

or sybsecurity databases), on a server running 15.5.x Cluster Edition, to be loaded into

a database on a server running 15.5.x non-Cluster Edition. This can be disabled using trace

flag 3411.

indexed encrypted column fails with an infected with 11 in stu_bincharinter

This however should not be the case.

1) If 'max parallel degree' > 1, then please disable parallel sort with

sp_configure 'max parallel degree', 1

or

2) For upgrade a database via dump and load, set dboption 'select into/bulkcopy/pllsort' to false for the database on the target server before load an older version of database dump:

sp_dboption db_name,'select',false

go

use db_name

go

checkpoint

go

load database db_name from ......

Provided the * expands to a single column in a single table/view/derived table:

1. the asterisk can be qualified with a table name (tab.*)

2. the subquery can be a GROUPBY query.

It will function as SELECTing the single column explicitly.

buffer is destroyed.

the processes of termination.

This new feature adds support to allow the alternative greedy algorithm to be used when the index row for a 12.5 GROUP BY key is too wide for the query to be executed in 12.5. This new feature can be activated by trace flag 16929 or optcriteria allow_large_index_row.

Also, when partitioned tables are involved in the query the alternative greedy algorithm will now provide an initial starting point using global statistics from the partitioned tables, but otherwise treating the partitioned tables as unpartitioned. This new feature can be activated by trace flag 15366 or optcriteria basic_optimization_partitions.

of a given database is changed by sp_logiosize, then Adaptive Server may incorrectly

ignore it during startup.

gap is observed after reboot.

when executing a query on a view defined on other nested

views which require a Worktable to materialize some

intermediate results and involve a join with mixed

datatypes.

With statement cache off, or literal autoparam off, the query chooses the index.

can no longer be dumped due to user '<username>' (suid <suid>) executing <cmdname>"

is now reported in the error log when the transaction dump sequence is

broken.

number of IN values exceeds 128. Consequently, when the number of such IN values exceeds

32K, signal 11 and stacktrace could happen in GaGenericArray::_GaGetPtr() when optimizer is

trying to create more than 32K (int16) internal predicate objects.

'optimization timeout limit' . This can be misleading. This request as to allow statement

cache lwps to use the 'optimization timeout limit' as per ad-hoc sql.

affects AIX platform only in both ASE SMP and Cluster editions running in

IBM AIX LPAR environment.

The new behavior is that SySAM obtains the CPU core count from the Entitled Capacity

in a Capped LPAR environment, and SySAM obtains the CPU core count from the Maximum

Capacity in an Uncapped LPAR environment.

Prior to ASE 15.5 ESD#3, SySAM obtains the CPU core count from the Maximum Virtual

CPUs and Active Physical CPUs in Capped and Uncapped LPAR environments.

duplicate row into the table does not raise error 2615.

LeScrollCache::LeScrCacheGetRow() when fetching rows from an insensitive cursor of a

select..union statement

occur when a client does a query.

'lt_sqlrep_comp_policy' together with a stack trace which

includes the modules 's_compile_stmt' and 'GenLava' may be

reported in the error log when a session executes a query that

is being compiled by ASE, while a REORG REBUILD command or

DDL activity takes place concurrently on the same table.

This CR introduced a new parameter to sp_droplogin (@with_override).

It is required to be set to 1 if wanting to drop a login when one or more databases

are unavailable to check login references.

datapartitions is not guaranteed in the same order of semantic partitioning boundary.

Fastfirstrow scenarios include min/max optimizations, exists

scans and the fastfirstrow optgoal. This can lead to

discarding fastfirstrow strategies in preference to other less

efficient strategies incorporating sorts. The fix for this is

enabled by the cr646024 optcriteria.

on ASE 15.5 with the error 1009, Level 16, State 2.

This is a request to change that behaviour and deny that command on an offline database.

Error 13968

Severity

16

Message text

"ALTER TABLE '%.*s' failed. You cannot drop all partitions of a table."

Explanation

The alter table command provides a method to remove a partition from table. Any data that is stored on that partition is moved to a suitable remaining partition.

When there is only one partition left on a table, it is no longer possible to store the data in an alternative partition. If an attempt is made to remove the final partition from a table using the ALTER TABLE command then error 13968 will be reported.

Action

This error is reported when an attempt is made to drop the only remaining partition of a table. If the requirement is to drop the table from the database then the correct method is to use a drop table command. If this is not the requirement then no action is required.

Additional Information

Refer to the Reference Manual: Commands for information about the alter table command

because upgrade found no entry in syscomments for this

object. Please drop and recreate dbo." may be reported

when executing a stored procedure which belongs to a

procedure group that was created using deferred name

resolution and at least one of the procedures in

the group has a SELECT * statement.

turned on does not insert the data.

in ASE 15.0. Cases include joins between partitioned and unpartitioned tables. This

change adds capabilities and costing support for these queries. Changes are enabled

using trace flags 16970 and 15331. You may also need to sp_chgattribute to set the plldegree on a per table basis.

recompiled after execution starts, a stack trace containing the functions, copyconst(),

copyres() and s_recompile() could be generated.

in a user defined temporary database may sleep for ever while ASE generates the

object ID value for this table. The call stack for the sleeping task will report

the modules 'crt_main' and 'objd_id'.

database 'db_name' (id), object 'table_name' (id), index 'index_name'

(255),..." or a 14108 error, "Could not find partition descriptor

for objid <...> ..." together with a stack trace which includes the

modules 'get_txtpage' and 'instext' may be reported in the error log

when running a SELECT INTO UNION which requires to use an internal

worktable (for example, because of an ORDER BY clause) and a NULL

literal value is used in one of the UNION sides for a TEXT, UNITEXT

or IMAGE column.

no open transaction.

Instead 0 was found." together with a stack trace which includes the modules 'des_unkeep'

and 'onl__chk_indexes' may be reported in the error log when a database dump is loaded

that has been previously obtained on a different platform.

reads the missing (not found the in ASE cache) pages from the disk and will never keep them in

the ASE cache. This incurs lot of i/o on large databases (database having lot of tables) and

may degrade the response time.

the module 'pci_configured' together with a stack trace which includes

the modules 'sm__exec_memload' and 'shm_memload' may be reported in

the error log when a memory dump is loaded with Sybmon while attached

to ASE server.

SQL statement as a parameter where a TEXT value is expected, implicit conversion will take

place within ASE. If the UNITEXT value is larger than 450 bytes and the converted TEXT

value is smaller than 450 bytes, a stack trace including the functions, memtextlen()and

LeLOBCache::_copyConstantToCache() will be generated.

from a Jconnect client as DBisql.

monErrorLog just after the ASE is restarted if the ASE is configured with multiple engines.

involving module ssql__rmplans(), ssql_save_SQL() and sqt_put_in_stmtcache().

sort operator costs. This problem is exasperated in parallel

query optimization.

columns even with the same default value, leading to unnecessary large number

of default objects in ASE.

any memory at run time, sp_helpconfig will raise error when estimate the memory usage on

them.

to reuse object descriptors whose statistics metadata are not yet flushed

on disk. The configuration parameter 'number of open objects' would need

to be increased accordingly.

and fail to detect that the log has been extended. The problem is more likely to be seen when

there is little activity and the transactions are short. RepAgent wakes up and resumes replication

once the log grows past the current last log page where it stalled.

Linux 64bit(both IBM P-series, x86_64 versions) is

((disk i/o structures + max async ios per engine+1) * number of engines)

into an ASE server is not found when required by executed java code.

Chapter 3: Using Templates to Create Jobs

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

Installing the templates into Job Scheduler

On UNIX:

installTemplateXML servername machinename serverport username password language

On Windows:

installTemplateXML.bat servername machinename serverport username password language

where:

In fact, the name of the template xml script is installTemplateXml (not installTemplateXML) in the Sybase file directory.

'des_unkeep_or_drop' or s_free() with a stack trace that includes the modules

'apm__metrics_insert', 'kill_proc' and 'proc__dealocmem' may be reported in the

error log after a session that was hanging because of the transaction log

being full was killed and the configuration option "enable metrics capture"

was active. This error resulted in ASE shutting down itself.

------

TF 758 enables this fix. It needs to be used at startup.

Start the server with -T758 and this will enable large scale allocations to use the ELC directly, i.e. 2 page\4 page\8 page.

a 2736 error may occur during recompiling a light-weight procedure if statement cache

is enabled and the user is not added to the current database.

"For a select query, the allowable limit is still 50 tables.

The total number of user tables allowed in a query is 512, you can only achieve that by using UNION, with 50 as the limit at single side of UNION."

0x090000000000d6c8 (pci_cfgmgr_default_config+0x7b0).

a non-coordinator instance, is performed serially so that

making all databases available takes an unnecessarily long period of time.

The opening of databases is now done in parallel. The previous

behavior has been retained under the 2796 trace flag.

together with a stack trace including the modules 'bufpindirty' and 'pg_allocate'.

Alternatively, the "infected with 11" message may be seen in the error log together

with a stack trace showing the modules 'xls_bufunpin' and 'buf_get_latch'.

Big integer request to have a trace flag to revert the behavior.

upgrade step [ID 53]

together with a stack trace which includes the modules 'pg_allocate' and

'bufpindirty' may be reported in the error log when a table was in the

process of being created and ASE has previously hit a 1204 error "ASE has run

out of LOCKS. Re-run your command when there are fewer active users, etc.".

when running alter database command if systhresholds system table is corrupted and

diagserver is used.

<value> etc." may be reported when an UPDATE with OR clauses is executed

on a Data Only Locked table while concurrent DML activity takes place on it.

The execution plan would report "Using Dynamic Index" and "Positioning by

Row IDentifier (RID)". Alternatively the message "The current row ID

'<value>' does not match the new row ID '<value>' for update, etc."

may be reported in the error log.

Queries that do not save the result set in a temporary table are no affected.

database with headeronly is issued from the master database for a user database and the source was

configured with a different character set and/or sort order.

when ALTER DATABASE and QUIESCE DATABASE are executed by two sessions concurrently.

when executing a stored procedure updating temporary tables

In the example below we only see the query type now, but we want to see the query plan from the select statement

SESSION 2 � Long running query

1> select @@spid

2> go

------

371

1> declare C1 cursor for select count(*) from

tempdb..sysobjects a,tempdb..sysobjects b,tempdb..sysobjects c

2> go

1> open C1

2> go

SESSION 1 � run showplan to see the query plan

1> sp_showplan 371, null, null, null

2> go

QUERY PLAN FOR STATEMENT 1 (at line 1).

STEP 1

The type of query is OPEN CURSOR C1.

(return status = 0)

Before opening the cursor, turn on set showplan, when the cursor is opened the plan is displayed.

declare C1 cursor for select count(*) from tempdb..sysobjects a,tempdb..sysobjects b,tempdb..sysobjects c

go

set showplan on

go

open C1

go

QUERY PLAN FOR STATEMENT 1 (at line 1).

STEP 1

The type of query is OPEN CURSOR C1.

QUERY PLAN FOR STATEMENT 1 (at line 1).

STEP 1

The type of query is DECLARE CURSOR.

5 operator(s) under root

|ROOT:EMIT Operator (VA = 5)

|

| |SCALAR AGGREGATE Operator (VA = 4)

| | Evaluate Ungrouped COUNT AGGREGATE.

| |

| | |N-ARY NESTED LOOP JOIN Operator (VA = 3) has 3 children.

| | |

| | | |SCAN Operator (VA = 0)

| | | | FROM TABLE

| | | | tempdb..sysobjects

| | | | a

| | | | Index : ncsysobjects

| | | | Forward Scan.

| | | | Positioning at index start.

| | | | Index contains all needed columns. Base table will not be read.

| | | | Using I/O Size 2 Kbytes for index leaf pages.

| | | | With LRU Buffer Replacement Strategy for index leaf pages.

| | |

| | | |SCAN Operator (VA = 1)

| | | | FROM TABLE

| | | | tempdb..sysobjects

| | | | b

| | | | Index : ncsysobjects

| | | | Forward Scan.

| | | | Positioning at index start.

| | | | Index contains all needed columns. Base table will not be read.

| | | | Using I/O Size 2 Kbytes for index leaf pages.

| | | | With LRU Buffer Replacement Strategy for index leaf pages.

| | |

| | | |SCAN Operator (VA = 2)

| | | | FROM TABLE

| | | | tempdb..sysobjects

| | | | c

| | | | Index : ncsysobjects

| | | | Forward Scan.

| | | | Positioning at index start.

| | | | Index contains all needed columns. Base table will not be read.

| | | | Using I/O Size 2 Kbytes for index leaf pages.

| | | | With LRU Buffer Replacement Strategy for index leaf pages.

the monitor counters for spinlocks may indicate that the DES manager spinlock was excessively used.

involving modules open_lwpbyname() and open__byname().

may be raised when an engine is offline and dbcc monitor is executed multiple times through sp_sysmon.

abort. This may cause an infinite loop of Error 1105, Msg 140030 or other errors.

's_retparams' together with a stack trace which includes the modules

's_setreturns' and 'execproc' may be reported in the error log when

a SELECT statement is run on a table having an encrypted column and

the configuration parameter "statement cache size" is greater than 0.

together with a stack trace involving the module tdsrecv__dynexec_xoltp() in case of recompilation.

partitioned tables degenerates to a single partition through dynamic

partition elimination and there is a CIS table in the query, then a SEGV could

occur in the PartNodeInfo::collocatedSite() method.

log segment is mixed with data segments may have inaccurate free space

counts.

"Procedure cache exhausted before a query plan could be found"

wide parameters that the number of bytes necessary to log an invocation would require

more than a single log page. The error will only occur if the stored procedure has

been previously invoked with null values in the wide parameter(s) and is

subsequently invoked with non-null values in those same parameter(s).

right information with regards to the hash by name table

of DEScriptors.

was in standby_access mode.

by executing the sp_config_rep_agent system stored procedure,

the password is logged encrypted to avoid it from being exposed

in clear text in the transaction log and during the replication

of this stored procedure execution.

deadlocked condition with another task which is truncating the transaction log.

into sharable inline default objects and drop the remaining duplicate ones.

<environments>

Sun Solaris UltraSparc-T / Solaris 10 / Doesn't configure Logical Domains.

<WorkAround> Use 1 or 2

1) use SYBASE_SAM_CAPACITY enviroment ( sysamcap DEFAULT )

2) add solaris.ldoms.read role to sybase user.

In addition, on a High Availability environment, when a primary fires a command to CREATE/ALTER/DROP login or login profile and if the command fails on the secondary, the TDS return

status of the command on the primary is still a success.

Installation Guide Adaptive Server Enterprise 15.5 for Sun Solaris

Installation Guide Adaptive Server Enterprise 15.5 - Cluster Edition for Sun Solaris

Preinstallation Tasks for the Server Installation :

wrong: project.max-shm-memory=\(privileged,17179869184,deny)

correct: project.max-shm-memory=(privileged,17179869184,deny)

DBConsole, and ML Monitor. To do this, you must deploy an OEM.ini file

along with DBISQL (et al). The file must contain the following lines:

[preferences]

directory=preferences_files_directory

where "preferences_files_directory" is a fully-qualified directory name,

e.g. "c:\work\prefs". The directory name should not end in a separator

(backslash on Windows; forward slash on Linux, Unix, and OS X).

Preferences files include:

.isqlPreferences12_32

.isqlPreferences12_64

.isqlHistory12

.jlogon12

.textCompleter12

.SybaseCentralEditor610

.scUserPreferences610_32

.scUserPreferences610_64

among others.

message together with a stack trace showing module name 'dbt_rdump_action_pcm_callback'. This

may happen after a DUMP DATABASE command was interrupted or hit an exception.

number=<page number> allocation status=<status value>", followed by error 7968, "The object

allocation map for tabid <object id> indid <index id> partition <partition nid> is corrupt,

therefore the optimized report cannot be generated. Contact Technical Support for assistance.",

will be reported when running "dbcc textalloc" with "noreport" option multiple times.

case of a MLD database when another task is doing the allocation on

the same allocation page or with any database if the buffer for the

new page could not be allocated.

may be reported when running SP_HELP_REP_AGENT with no parameter values.

together with stack trace involving modules s_reinstall_lwp_tree(), s_procinstall_lwp(),

s_recompile() and tdsrecv__dynexec_xoltp().

DistinctHashing operator code generation when there are

two distinct expression columns and one is becoming dependent on

the other through equijoin transitive closure. Using diagserver,

the assertion "Assertion (GcvIsVirtualColumn(pNode))" could fail

in cgevals.cpp line 202.

'fglk__send_prom_attn' together with a stack trace which

includes the modules 'fglk__promote' and 'getnext' may be

reported in the error log and may bring the server down

if a query involving a big number of tables (more than 50)

was being executed in parallel and due to the number of

locks acquired it attempted to promote the locks to a

lock of a higher level.

VTABRemoteAccess::getNext() during a period of heavy user activity.

inner to an outer join is referenced in

a higher scope outer join or in the WHERE

clause, it would cause the outer join to

degenerate into an inner join since all

null extended rows would be eliminated. The exception

occurs with "null preserving" predicates such

as IS NULL which would not eliminate the row.

If a nested outer join exists where

a table is inner to an outer join and

an attribute of the table is referenced in second

higher scope outer join ON clause in a "null preserving"

predicate then a wrong answer could result when a mergejoin/hashjoin

is chosen for the higher scope outer join.

sysstatistics. A subsequent UPDATE STATS reduces the number of steps to an expected value.

(1) a table owner who has decrypt permission on an encrypted column through group membership

is not permitted to select decrypted data from the column.

(2) sso_role who should have implicit decrypt permission is not permitted to select decrypted data from the column and is not permitted to insert into the column when trigger present on the table

(3) a table owner who has been granted decrypt permission through a user defined role is not permitted to insert data into the column when trigger present on the table

together with a stack trace which includes the modules 'dsinit' and 'rec__backout'

may be reported in the error log during ASE startup if a 708 error, "There is

insufficient system memory to run kernel task. spid <value>, kpid <value>."

was previously reported in the error log.

password is provided as a parameter on the command line.

for a non-ha server the following message will be issued by the companion server:

Msg�10330,燣evel�14,燬tate�1:

Server 'gree1_22760',燩rocedure 'sp_attrib_notify':

EXECUTE爌ermission燿enied爋n爋bject爏p_attrib_notify,燿atabase爏ybsystemprocs,

owner燿bo

the module Partitions::initHintsForDegree() may be reported when a remote procedure is executed with the recompile option.

transaction and command rollback.

together with a stack trace which includes the modules 'pg_allocate' and

'bufpindirty' may be reported in the error log when a table was in the

process of being created and ASE has previously hit a 1204 error "ASE has run

out of LOCKS. Re-run your command when there are fewer active users, etc.".

'rec_undo_session' while accessing the buffer address from an address lock of an APL page.

with UNION ALL can improve the compilation performance in ASE 15.x.

Adaptive Server Enterprise version 15.7 ESD#4 and later now support the UNIX domain socket

communication, which avoids some of the overhead the TCP protocol incurs.

Similar to other protocols, include the AF_UNIX protocol in the interfaces file for clients to

connect to Adaptive Server. The syntax for an AF_UNIX in the interfaces file is:

master afunix unused //<hostname>/<pipe>

<number> and <number> at offset <number>..." may be reported in the

errorlog when a session executed a trigger which contains a SELECT FOR XML

referencing INSERTED or DELETED virtual tables and previously the same

session had been executed a different trigger for a different table

but with the same SELECT FOR XML statement.

increasing the loads. Requires to increase heap memory per user.

a column, executing DBCC TEXTALLOC on the table which still have some TEXT/UNITEXT/IMAGE columns will report 15918 error.

not reported in the estimated cost diagnostics (statistics

plancost) for parallel optimized strategies.

relative pathname that includes '../'

Bridge directive PCI_BRIDGE_X_OPT." followed by the message "current process infected with 11"

in the module 'pci__cfg_plb_options' together with a stack trace which includes the modules

'dsinit' and 'pci_cfgmgr_init' may be reported in the error log during ASE startup whose

logical page size is greater than 8K if ASE first reported an error during the recovery

of a user database and the configuration option 'enable pci' is active.

if max_rows_per_page is set to 1 for the non-clustered index on APL table.

In a XML model, the Orientation display preference options are misaligned.

To reproduce:

1) open a XML model

2) select Tools\Display Preferences

3) Select General setting and General tab

Note: for the Orientation section, the Horizontal option appears to centered vertically and the Vertical option appears to be set to 'Bottom'