2081320 - Targeted CR List for ASE 15.5 ESD#5.2 - SAP ASE

CR number

Description

588137

Duplicate key values could be generated for identity column during checkpoint of the database. For tables with unique indexes on identity column, it would give error 2601 'Attempt to insert duplicate key row in object < obj_name > with unique index < index_name > '. more co A more complete version of this fix has been developeed under CR 614823, i.e. 614823 supersedes 588137.

610956

ASE installations with large number of databases, with high number of engines and inter database transactions would result in high contention of dbt_spin.

622538

In rare circumstances, a 12323 error may be raised when running REORG REBUILD on nonclustered index of a DATA_ONLY LOCKED table.

636786

The execution of SP_RENAME and SP_CONFIGURE will raise error 17260 "Can't run < procname > from within a transaction." incorrectly when invoked under chained transaction mode despite that there is no open transaction.

662410

ARIES CE with > 1 instance setup hits SIG11 inside function lock__mmaster_processque

663600

Updating identity columns gives error 586: Illegal identity value specified for identity

665715

The message "current process infected with 11" in the module 'bufdlink' 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.".

669847

A 290024 error, "Invalid column length: < number > Value must be between < number > and < number > at offset < number > ..." may be reported during a UPDATE plan execution when a complex VIEW or derived table (such as UNION or GROUPBY) is used in the FROM clause of the UPDATE statement.

672446

After entered the ASE login/password in the login dialog to configure the existing ASE, syconfig crashed.

674627

When a table is being replicated to a table that includes an additional IDENTITY, an error "AOP not found in getAopt" is raised.

676148

The ASE PCI layer responsible for the execution handshake of ADT and UDF Java operations with the integrated COTS JavaVM is enhanced to achieve better performance.

676991

High cpu usage seen with rprocmgr_spin contention

680759

In rare circumstances, terminating an idle CIS connection may result in some form of memory corruption in the next connection attempt.

680776

When using some devices (i.e. EQLOGIC) on Linux with multipathing, ASE cannot correctly locate the alternate paths to the raw devices for I/O fencing which may prevent ASE from starting up.

681948

The threshold status passed to the threshold stored procedure sp_thresholdaction may contain some unexpected data for thresholds other than the last chance threshold.

682738

CIS creates the first text page in an INSERT-SELECT query even though the source data is null. To activate the fix, trace flag 11241 needs to be turned on.

683626

Abstract plan error message could be reported when using "set forceplan on" for outer join query with subquery.

685996

Abstract plan error could happen failing to apply a right nested outerjoin abstract plan. It is more likely to happen when there is a subquery at the inner side too. Same issue could also cause optimizer to ignore certain legit merge/hash outerjoin bushy plan too under allrows_dss optimization goal.

687303

We need a code change for histserver to allow the increase of the SRV_S_DEFQUEUESIZE for customers who use histserver against multiple ASEs. The current default is not sufficient.

687523

In rare circumstances on a heavily-loaded multi-engine system, a 8201 error "Keep count of descriptor (objid= < id > , dbid= < id > ) was expected to be 1. Instead 0 was found." may be reported in the error log together with a stack trace which includes the modules 'des__unkeep' and 'close_range' when an INSERT, DELETE or UPDATE is executed on a table that has many referential integrity constraints defined and after the execution of one of its triggers. Alternatively the message "infected with 11" in the modules 'seq_schema_check_virtab' or 'LeScanOp::_LeOpOpen' may be reported.

690064

A 11051 error, "ALTER TABLE ' < table_name > ' failed. Table is currently being used by one or more other tasks." may be reported when a REORG REBUILD command is run while another process concurrently runs a SELECT query on the same table at isolation level read uncommitted. Following this error and while the SELECT query continues to execute, any process that attempts to reference this table will fail mistakenly with an error such as 17461 "Object does not exist in this database.".

690406

When the coordinator node is shutdown, the new coordinator may be fenced out when new instance joins the cluster. The new coordinator will display write errors and then shutdown e.g. "sddone: write error on virtual disk 0 block 4598". This issue has been resolved via CRs 680776 and CR 686131.

690730

High rprocmgr spinlock contention may be observed under heavy workload of dynamic SQL and stored procedures.

691248

When tempdb_space resource limit is reached, any subsequent work table creation operation fails and does not release the buffers and the allocated tempdb pages during the execution.

692333

In rare circumstances on a heavily-loaded multi-engine system, the message "timeslice < value > , current process infected" together with a stack trace which includes the modules 'jldr_ReleaseLocalLoader' and 'pci_free' may be reported in the error log when applications use JAVA extensively.

692872

In rare circumstances, Adaptive Server may hang as a result of an undetected deadlock between a process that is executing a DUMP TRANSACTION and another process that is running a query that uses the last-chance threshold builtin LCT_ADMIN.

694961

Reservation conflict messages may appear in the system log on systems with multipathing when I/O fencing is enabled.

695156

When ASE is configured with multi-bytes charset with 'statement cache' and 'literal autoparam' enabled, repeatedly executing a SQL batch with comments containing multiple byte characters can cause error 102.

696837

In a 2 node or greater ASE Cluster Edition setup, a database with a log segment threshold defined may result in incorrect values for unreserved data pages and no thresholds enabled. The database will report log is full and transactions will either suspend or abort, depending on the configuration paremeter "abort tran on log full" value. This can only occur when the database has multiple thresholds and the second threshold is crossed before the first one is completely handled.

697828

High rprocmgr or rpmctrl spinlock contention may be observed under heavy workload of dynamic SQL or stored procedures.

698063

When an INSERT query attempts to insert an explicit value for a TIMESTAMP column a warning will be reported, but query execution continues. If the explicit value for the TIMESTAMP column is the result of evaluating an expression, and the query plan for the INSERT is re-executed (i.e. because the statement cache and literal auto parameterization are enabled or because the INSERT query is part of a stored procedure), memory corruption may occur, leading to various stack traces.

698134

In some rare scenarios, procedure cache memory tracking data structures may get accessed after they are freed to global memory queue. This may lead to procedure cache memory corruption.

698315

The message "current process infected with 11" in the module 'bufpindirty' together with a stack trace which includes the modules 'xact__rollbackxact' and 's_execute' may be reported in the error log if a SELECT INTO statement involving a big number of tables (more than 96) is aborted with a 226 error, "SELECT INTO command not allowed within multi-statement transaction". If after that a ROLLBACK TRAN is executed by the same session a 6203 error "Unable to do cleanup for the killed process; received Msg ..." may also be reported in the error log.

700404

Optcriteria 619756 (enabled in optlevels ase1503esd4 or ase_current) could generate a SEGV and/or assertion failure. The stacktrace associated with the SEGV will typically include references to optimizer routines. In a diagserver, the following assertion (eqc- > eqcIsUnrefed() || OptDuringGtuLocalGroupingPrimingHack()) may fail.

700652

In rare circumstances, a signal 11 may be reported in the errorlog with a stack trace involving the sort module if a reformatting query plan is used.

701154

The message, "timeslice -501, current process infected at 0x < address > ", in the module 'cm_writelogdes_upto_sequence' together with a stack trace which includes the modules 'cmcc__wri_bcmt_proc' and 'bufwritedes_log' may be reported in the error log when ASE flushes a syslogs page on disk while downgrading the physical lock held on this syslogs page in a 2 node Cluster Edition setup.

702653

When a primary key contains IDENTITY columns, INSERT INTO and SELECT statements may create bad LTL causing the RepServer to report the error message, "Incorrect syntax near the keyword FROM". This is fixed via CR 674627.

702770

Dataserver might hit SIGSEGV in is_vpar() during startup on HP PA-RISC vPar.

703123

Predicates that contain more than 128 or-terms in a disjunct may not be optimized using a general-or strategy. In cases where such a general-or strategy provides the only efficient search strategy, query performance could be poor.

709510

Enabling I/O fencing may result in "reservation conflict" messages in the system log.

712855

Invoking the soundex() or difference() built-ins on unichar/univarchar strings larger than 127 characters (254 bytes) may cause the process to be terminated.

712903

Idle connections may fail to migrate during instance shutdown.

716870

Adaptive Server may not kill a task that is marked for termination, which is reporting shortage of buffers in buffer pool like "cache is configured too small for current demands".

717079

The ASE COTS JavaVM version is updated to version JRE 6 update 33 for each supported platform in the sybpcidb database.

718191

Error 2812 or 11214 is raised when accessing an MDA table specifically mapped as a proxy table.

718637

Parallel queries with the following query plan scenario MIGHT fail with a "Command has been aborted" message and no other warnings/errors reported in the error log: o There is a merge-join op where one of its child trees contains a replication exchange operator. o There is no "draining" operator (e.g. sort operator) between the merge-join and the exchange operator. o The merge-join's other child drains its input stream before the child with the replication exchange does.

719449

Adaptive Server may encounter a storage access violation error when query execution is interrupted or some special query is executed on 64bit diagsrvr.exe or 32bit sqlsrvr.exe on windows platform.

719673

A 14108 error, "Could not find partition descriptor for objid %d, indid %d, ptnid %d in dbid %d" may be reported when running DBCC DBREBOOT if traceflag 12315 is off.

720659

Very slow optimisation of queries with large IN clauses Excessive stack usage with aggregate queries with large IN clauses

721144

If Java in the database is being used, then if the "same" shared configuration file is used by more than one instance or Server, then there is a race condition during JVM initialization where it is possible that the message ?current process infected with signal 11? in the function ?jvminit? may be reported in the error log, followed by the message ?Failed to initialize Java services". This may result in ASE shutting itself down.

721959

Literal auto parameterization of large IN lists may cause performance issues. Trace flag 7786 was introduced to turn off literal auto parameterization of queries with large IN lists ( > 128 constants) as well as avoiding caching of the respective statement. Trace flag 7786 only has an effect if literal auto parameterization is turned on i.e. if literal auto parameterization is turned off, then the statement will still be cached.