Re: Log switches every 4 minutes with 17MB Archivelogs despite 500MB log file size

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Wed, 1 Feb 2017 11:32:11 +0000
Message-ID: <MMXP123MB0911ABD9E1D599FD8B72C55BA54D0_at_MMXP123MB0911.GBRP123.PROD.OUTLOOK.COM>


You didn't answer my questions about frequency and whether it's both nodes or just one. Since you're still getting archived logs which are much smaller than the redo log size I doubt if you've got a "solution", only a change in timing.

Were the small archived logs actually causing a problem ?

Regards
Jonathan Lewis



From: oracle-l-bounce_at_freelists.org <oracle-l-bounce_at_freelists.org> on behalf of Martin Decker <martin.decker_at_ora-solutions.net> Sent: 01 February 2017 10:04:06
To: oracle-l_at_freelists.org
Subject: Re: Log switches every 4 minutes with 17MB Archivelogs despite 500MB log file size

Dear all,

there is no dataguard standby involved. No Redo Shipping. No DG Broker.

Oracle Support finally came up with a solution: After increasing redo log files from 512M to 1024M, we no longer see redo log switches every 4 minutes but only after 50 minutes and the archived redo logs are no longer only 15-17MB in size but around 180MB. Of course this is documented neither in the official documentation nor on MOS.

Unfortunately, they could not give a plausible reason for the behaviour.

Another colleague mentioned that the huge 512M granule size could also have an effect. We are planning to reduce KSM granule size to 128M at the next maintenance window.

We are also discussing to adapt some other RAC underscores as described in this Note but have not decided yet: Best Practices and Recommendations for RAC databases with SGA size over 100GB (Doc ID 1619155.1)

Best regards,
 Martin

2017-01-31 17:35 GMT+01:00 Matt Adams <MAdams_at_equian.com<mailto:MAdams_at_equian.com>>: That phrase ‘MAXIMUM PERFORMANCE’ in the log file makes my thing that standby database management might be involved, but I don’t think I’ve ever seen quite this situation before.

We have standby databases, do not use the Data guard broker.

Is this rac cluster also the primary for a standby site? Is it possible that the data guard broker software is doing this somehow to force changes from a primary to a standby database?

From: oracle-l-bounce_at_freelists.org<mailto:oracle-l-bounce_at_freelists.org> [mailto:oracle-l-bounce_at_freelists.org<mailto:oracle-l-bounce_at_freelists.org>] On Behalf Of Martin Decker Sent: Tuesday, January 31, 2017 9:23 AM
To: Powell, Mark
Cc: oracle-l_at_freelists.org<mailto:oracle-l_at_freelists.org> Subject: Re: Log switches every 4 minutes with 17MB Archivelogs despite 500MB log file size

Dear Mark, Bertrand,
log_checkpoint_interval is not net
log_checkpoint_timeout is not set
archive_lag_target is not set
no cron jobs
no dba_jobs
no dba_scheduler_jobs
"alter system" audit enabled, but no audit records generated in neither audit_file_dest nor dba_audit_trail fast_start_mttr_target was not set, and is now set to 300s, but does not a difference. Redo Log Buffer is not specified and defaults to around 500 MB, we have 4 CPU Cores with hyperthreading, so 8 visible cores per host.

More details (v$log, init.ora contents, x$kcrfstrand) can be found here: https://drive.google.com/file/d/0B_d8Sdgtmy-rUGlIWkFFaWJhYWc/view Regards,
Martin

2017-01-31 14:54 GMT+01:00 Powell, Mark <mark.powell2_at_hpe.com<mailto:mark.powell2_at_hpe.com>>:

What are the values database parameters log_checkpoint_interval and log_checkpoint_timeout?



From: oracle-l-bounce_at_freelists.org<mailto:oracle-l-bounce_at_freelists.org> <oracle-l-bounce_at_freelists.org<mailto:oracle-l-bounce_at_freelists.org>> on behalf of Martin Decker <martin.decker_at_ora-solutions.net<mailto:martin.decker_at_ora-solutions.net>> Sent: Tuesday, January 31, 2017 8:05:55 AM To: oracle-l_at_freelists.org<mailto:oracle-l_at_freelists.org> Subject: Log switches every 4 minutes with 17MB Archivelogs despite 500MB log file size

Dear List,
in a RAC 2 Node cluster (12.1.0.2 with DBBP October 2016), we see redo log switches every 4 minutes with Archivelogs filled up to 15-18 MB. Redo Log File size is 512MB. The Application is using the instances with singleton services so node1 has much more redo generation compared to node2. It is unclear what is triggering the redo log switches. We can exclude the following:

* ALTER SYSTEM SWITCH LOGFILE;
* ALTER SYSTEM ARCHIVE LOG CURRENT;
* RMAN ARCHIVELOG / DB BACKUP
* ARCHIVE_LAG_TARGET

Oracle Support SR is open, but so far no clues. Any ideas on how to diagnose? LOG_ARCHIVE_TRACE shows that log switch is exectued, but not who/what is triggering the switch:
  • 2017-01-19 14:13:58.496248 1380 krsl.c krsl_log_switch_trace: Beginning Log Switch operation for MAXIMUM PERFORMANCE: krsk_ler_get: Read LE 16 flag 0xa T-1.S-888 low 0x000c.d4c16698 nxt 0xffff.ffffffff nab 4294967295 switching from log 16 thread 1 sequence 888 to log 12 thread 1 sequence 889 Best regards, Martin

--
--

Martin Decker
Oracle 10g/11g/12c Certified Master
+43 (0) 699 1131 6463<tel:+43%20699%2011316463>

martin.decker_at_ora-solutions.net<mailto:martin.decker_at_ora-solutions.net> http://www.ora-solutions.net

--
--

[Image removed by sender.]

Martin Decker
Dipl.-Ing. (FH)
Oracle 10g/11g/12c Certified Master

+43 (0) 699 1131 6463<tel:+43%20699%2011316463>

martin.decker_at_ora-solutions.net<mailto:martin.decker_at_ora-solutions.net> http://www.ora-solutions.net

[Image removed by sender.]

  • This communication may contain privileged and/or confidential information. If you are not the intended recipient, you are hereby notified that disclosing, copying, or distributing of the contents is strictly prohibited. If you have received this message in error, please contact the sender immediately and destroy any copies of this document. ****

--
--

[http://www.ora-solutions.net/images/ora-solutions_Mail3.gif]

Martin Decker
Dipl.-Ing. (FH)
Oracle 10g/11g/12c Certified Master

+43 (0) 699 1131 6463

martin.decker_at_ora-solutions.net<mailto:martin.decker_at_ora-solutions.net> http://www.ora-solutions.net

[https://docs.google.com/uc?export=download&id=0B_d8Sdgtmy-rOUtNN1laWTI0UWM&revid=0B_d8Sdgtmy-rY2lqbGx0S2JKWE5FUnk5QjVtSkR4NWFEeHRjPQ]

--

http://www.freelists.org/webpage/oracle-l Received on Wed Feb 01 2017 - 12:32:11 CET

Original text of this message