Oracle DBA – A lifelong learning experience

Posts Tagged ‘Updating SGAs in kzam_check_limit’

bug 8525592 – SGA memory leak during db shutdown

Posted by John Hallas on June 25, 2009

10 days ago I posted regarding informational messages in trace files and how they seem to have increased with the advent of 11G. See http://jhdba.wordpress.com/2009/06/16/informational-trace-files-in-11g/

I referred to 3 messages, one of which I said I was going to raise a SR against. The trace file contained the line  “Updating SGAs in kzam_check_limit”
and the full trace showed

 Updating SGAs in kzam_check_limit
No rows for property:16, trail_type:12
Resetting MaxSize to default values for OS audit files
Resetting MaxSize to default values for XML audit files
No rows for property:17, trail_type:12
Resetting MaxAge to default values for OS audit files
Resetting MaxAge to default values for XML audit files
No rows for property:24, trail_type:15

The feedback from Oracle was that this was unpublished bug  8525592 - SGA MEMORY LEAK ERROR DURING THE DB SHUTDOWN and that there was no cause for concern.

Now this was appearing on an 11.1.0.7 database on HPUX Itanium and it is the only instance out of around 25 where it is appearing. Co-incidentally we have a  memory leak on that server and we have an ongoing SR open with Oracle. I have alerted my colleague who owns that SR and he will talk to Oracle and see if there is any match between the two issues.

In the meantime I have been able to get a bit more information on what bug 8525592 is about. Apparently it is not a serious issue and isn't really a memory leak.  It's to do freeing up shared pool when there are multiple subpools.  If there are child/parent handles split between subpools, it's possible for the parent one not to be cleared properly.  In normal operation there's a houskeeping routine that SMON does periodically that sorts this out but if the database is shutdown before that runs, it looks like a memory leak.  It's only in 11.1.0.x and is not considered serious. Bug 8525592 leads me to bug 8586531 and base bug 7453413

Posted in 11g new features, Oracle | Tagged: , , , , , , | 1 Comment »

informational trace files in 11g

Posted by John Hallas on June 16, 2009

Since Oracle 11G there seems to be many more informational trace files generated than ever before. Perhaps I am noticing more but I will provide three examples in support of what I am saying.

kcrroda: calling ksfdrcres to create AL or RL

Just a warning but it is due to a bug in 11.1.0.6 (still not fixed in 11.1.0.7) that causes diagnostic warnings to be issued after an archivelog switch. Bug 6910132 for anyone who is interested

From the same database I see trace files with the following lines in

Updating SGAs in kzam_check_limit
No rows for property:16, trail_type:12
Resetting MaxSize to default values for OS audit files
Resetting MaxSize to default values for XML audit files
No rows for property:17, trail_type:12
Resetting MaxAge to default values for OS audit files
Resetting MaxAge to default values for XML audit files
No rows for property:24, trail_type:15

I cannot find any information on this at all but I do suspect that it is another diagnostic setting that has been left enabled within the code
I will raise an SR on this matter but I don’t think it is anything for me to be worried by

From another database (again 11.1.0.7) I see numerous trace files containing the line

qerfxGCol:KQFDTTIM – Error converting to LdiDateArray

Metalink note 577674.1 shows that it is yet another trace output than can be ignored
Applies to:
Oracle Server – Enterprise Edition – Version: 11.1.0.6
This problem can occur on any platform.

Symptoms
After upgraded to 11.1.0.6.0 or created a new database on this version, you noticed following repetitively message at MMON process trace:

“qerfxGCol:KQFDTTIM – Error converting to LdiDateArray”

There is no further errors at alert log.

Changes
Upgraded to 11.1.0.6.0.

Cause
This issue has been introduced in 11gR1 11.1.0.6.0 with the new view v$persistent_queues, where a ‘select * from v$persistent_queues;’ reproduces the error when there is a ‘null’ value for one of timestamp
columns in that view. This behavior does *not* cause any data loss; performance impact nor any effect for use of ADDM/AWR tools.

This issue has been reported on following bugs:

Bug 6780602 – TRACEFILES WITH ‘QERFXGCOL:KQFDTTIM – ERROR CONVERTING TO LDIDATEARRAY’
Bug 6921602 – QERFXGCOL:KQFDTTIM – ERROR CONVERTING TO LDIDATEARRAY

Solution
It has been fixed in 11.2.

Posted in 11g new features, Oracle | Tagged: , , , , , , , | 1 Comment »

 
Follow

Get every new post delivered to your Inbox.

Join 190 other followers