Fuzzy vs Full Checkpoints v10.00.UC6
Mike Magie asked why IDS 10.00.UC6 on Solaris 8 was logging mostly full "Checkpoint Completed" entries instead of fuzzy checkpoints, despite a 5-minute interval and no archives, log backups or conversions running. After others asked about HDR (none), ONCONFIG and I/O details, John Miller explained the cause: fuzzy checkpoints only apply to buffers belonging to logging databases, and since most of Mike's databases were non-logged, the logical logs couldn't be used for fuzzy recovery, so few buffers were left unflushed. Another poster suggested simply disabling fuzzy checkpoints (NOFUZZYCKPT) and tuning checkpoints instead.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
All, I am stymied as to why we see our engine performing full checkpoints as opposed to fuzzy checkpoints. In looking at the Admin Guide I understand the scenarios under which each type of checkpoint should occur. In our situation, based on the scenarios described we should not be seeing full checkpoints; no archive is occuring, no log backups, no conversion routines, etc. You can see the current log does not change during the time frame captured - most of our databases are non-logged. Our checkpoint interval is set to 5 minutes. There was moderate update activity occuring during the time frame captured, most of which would be considered fuzzy operations. There potentially could have been updates deemed non-fuzzy as well taking place. online.log excerpt (edited out lines containing " Maximum server connections 8) 09:43:40 Fuzzy Checkpoint Completed: duration was 3 seconds, 1 buffers not flushed. 09:43:40 Checkpoint loguniq 9949, logpos 0x6b050, timestamp: 0xc46e7fcc 09:50:07 Checkpoint Completed: duration was 87 seconds. 09:50:07 Checkpoint loguniq 9949, logpos 0x30f018, timestamp: 0xc48278a6 09:55:44 Checkpoint Completed: duration was 37 seconds. 09:55:44 Checkpoint loguniq 9949, logpos 0x52c018, timestamp: 0xc49b2bfa 10:01:12 Checkpoint Completed: duration was 28 seconds. 10:01:12 Checkpoint loguniq 9949, logpos 0x52f018, timestamp: 0xc4aaa99f 10:06:44 Checkpoint Completed: duration was 32 seconds. 10:06:44 Checkpoint loguniq 9949, logpos 0x13c7018, timestamp: 0xc4bbb441 10:12:07 Checkpoint Completed: duration was 23 seconds. 10:12:07 Checkpoint loguniq 9949, logpos 0x13c9018, timestamp: 0xc4cbdf1c 10:17:38 Checkpoint Completed: duration was 31 seconds. 10:17:38 Checkpoint loguniq 9949, logpos 0x13d5018, timestamp: 0xc4db9468 10:23:04 Fuzzy Checkpoint Completed: duration was 26 seconds, 4 buffers not flushed. 10:23:04 Checkpoint loguniq 9949, logpos 0x13dc080, timestamp: 0xc4e16301 10:33:37 Checkpoint Completed: duration was 33 seconds. 10:33:37 Checkpoint loguniq 9949, logpos 0x13dd018, timestamp: 0xc4e92e92 10:38:53 Checkpoint Completed: duration was 16 seconds. 10:38:53 Checkpoint loguniq 9949, logpos 0x13e4018, timestamp: 0xc4edef80
OS version? IDS version? Can we assume there is no HDR? HDR only does full checkpoints. Did this just start happening? Bob Roussey Unix / Informix Administration Spirit Airlines Robert.Roussey@SpiritAir.com -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of MIKE MAGIE Sent: Monday, December 17, 2007 2:38 PM To: ids@iiug.org Subject: Fuzzy vs Full Checkpoints v10.00.UC6 [10772] All, I am stymied as to why we see our engine performing full checkpoints as opposed to fuzzy checkpoints. In looking at the Admin Guide I understand the scenarios under which each type of checkpoint should occur. In our situation, based on the scenarios described we should not be seeing full checkpoints; no archive is occuring, no log backups, no conversion routines, etc. You can see the current log does not change during the time frame captured - most of our databases are non-logged. Our checkpoint interval is set to 5 minutes. There was moderate update activity occuring during the time frame captured, most of which would be considered fuzzy operations. There potentially could have been updates deemed non-fuzzy as well taking place. online.log excerpt (edited out lines containing " Maximum server connections 8) 09:43:40 Fuzzy Checkpoint Completed: duration was 3 seconds, 1 buffers not flushed. 09:43:40 Checkpoint loguniq 9949, logpos 0x6b050, timestamp: 0xc46e7fcc 09:50:07 Checkpoint Completed: duration was 87 seconds. 09:50:07 Checkpoint loguniq 9949, logpos 0x30f018, timestamp: 0xc48278a6 09:55:44 Checkpoint Completed: duration was 37 seconds. 09:55:44 Checkpoint loguniq 9949, logpos 0x52c018, timestamp: 0xc49b2bfa 10:01:12 Checkpoint Completed: duration was 28 seconds. 10:01:12 Checkpoint loguniq 9949, logpos 0x52f018, timestamp: 0xc4aaa99f 10:06:44 Checkpoint Completed: duration was 32 seconds. 10:06:44 Checkpoint loguniq 9949, logpos 0x13c7018, timestamp: 0xc4bbb441 10:12:07 Checkpoint Completed: duration was 23 seconds. 10:12:07 Checkpoint loguniq 9949, logpos 0x13c9018, timestamp: 0xc4cbdf1c 10:17:38 Checkpoint Completed: duration was 31 seconds. 10:17:38 Checkpoint loguniq 9949, logpos 0x13d5018, timestamp: 0xc4db9468 10:23:04 Fuzzy Checkpoint Completed: duration was 26 seconds, 4 buffers not flushed. 10:23:04 Checkpoint loguniq 9949, logpos 0x13dc080, timestamp: 0xc4e16301 10:33:37 Checkpoint Completed: duration was 33 seconds. 10:33:37 Checkpoint loguniq 9949, logpos 0x13dd018, timestamp: 0xc4e92e92 10:38:53 Checkpoint Completed: duration was 16 seconds. 10:38:53 Checkpoint loguniq 9949, logpos 0x13e4018, timestamp: 0xc4edef80 ************************************************************************ ******* Forum Note: Use "Reply" to post a response in the discussion forum.
10.00.UC6 - Solaris 8 You assume correct - no HDR. Dunno when it started happening - I just noticed it after a colleague asked me the question. The whole "issue" borders on esoteric but I was asked "Why?"... Mike
Reply with your ONCONFIG in the email (no attachments), # of real processors and RAM and some pretty wise people can make some recommendations. Bob Roussey Unix / Informix Administration Spirit Airlines Robert.Roussey@SpiritAir.com -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of MIKE MAGIE Sent: Monday, December 17, 2007 2:45 PM To: ids@iiug.org Subject: Re: RE: Fuzzy vs Full Checkpoints v10.00.UC6 [10774] 10.00.UC6 - Solaris 8 You assume correct - no HDR. Dunno when it started happening - I just noticed it after a colleague asked me the question. The whole "issue" borders on esoteric but I was asked "Why?"... Mike ************************************************************************ ******* Forum Note: Use "Reply" to post a response in the discussion forum.
I suggest to disable fuzzy checkpoints - set NOFUZZYCKPT environment variable before you start IDS. and do sensitive checkpoint tuning. In my opinion fuzzy checkpoints provide too little performance benefit (any ?) at the price of too many bugs (crash recovery not working). In IDS 11 fuzzy checkpoints are completely removed. Rgds Tilman
Sorry, some info. would also help - 1. Is database SAN connected? 2. Update Stats frequency? 3. onchecks being performed or not? 4. sar stats (I am assuming sar or similar also will be on Solaris), avg. service time values in particular? 5. Logical log backup on disk/tape, SM (ISM or any thrid party SM)? Cheers MIKE MAGIE <jmmagie@yahoo.com> wrote: 10.00.UC6 - Solaris 8 You assume correct - no HDR. Dunno when it started happening - I just noticed it after a colleague asked me the question. The whole "issue" borders on esoteric but I was asked "Why?"... Mike ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum. --------------------------------- Looking for last minute shopping deals? Find them fast with Yahoo! Search.
MIKE: Fuzzy Checkpoints will only be utilized for buffers in the buffer pool belonging to LOGGING databases. If the database is non-logging then the logical logs can not be used for "fuzzy recovery" of the buffers. = This is why you are see so few buffers not being flushed at checkpoint time. John F. Miller III = "MIKE MAGIE" = <jmmagie@yahoo.co = m> = To Sent by: ids@iiug.org = ids-bounces@iiug. = cc org = Subj= ect Fuzzy vs Full Checkpoints = 12/17/2007 11:37 v10.00.UC6 [10772] = AM = = = Please respond to = ids@iiug.org = = = All, I am stymied as to why we see our engine performing full checkpoints as= opposed to fuzzy checkpoints. In looking at the Admin Guide I understan= d the scenarios under which each type of checkpoint should occur. In our situation, based on the scenarios described we should not be seeing full checkpoin= ts; no archive is occuring, no log backups, no conversion routines, etc. You c= an see the current log does not change during the time frame captured - most o= f our databases are non-logged. Our checkpoint interval is set to 5 minutes. There was moderate update activity occuring during the time frame captured, most of which would b= e considered fuzzy operations. There potentially could have been updates deemed non-fuzzy as well taking place. online.log excerpt (edited out lines containing " Maximum server connections 8) 09:43:40 Fuzzy Checkpoint Completed: duration was 3 seconds, 1 buffers = not flushed. 09:43:40 Checkpoint loguniq 9949, logpos 0x6b050, timestamp: 0xc46e7fcc= 09:50:07 Checkpoint Completed: duration was 87 seconds. 09:50:07 Checkpoint loguniq 9949, logpos 0x30f018, timestamp: 0xc48278a= 6 09:55:44 Checkpoint Completed: duration was 37 seconds. 09:55:44 Checkpoint loguniq 9949, logpos 0x52c018, timestamp: 0xc49b2bf= a 10:01:12 Checkpoint Completed: duration was 28 seconds. 10:01:12 Checkpoint loguniq 9949, logpos 0x52f018, timestamp: 0xc4aaa99= f 10:06:44 Checkpoint Completed: duration was 32 seconds. 10:06:44 Checkpoint loguniq 9949, logpos 0x13c7018, timestamp: 0xc4bbb4= 41 10:12:07 Checkpoint Completed: duration was 23 seconds. 10:12:07 Checkpoint loguniq 9949, logpos 0x13c9018, timestamp: 0xc4cbdf= 1c 10:17:38 Checkpoint Completed: duration was 31 seconds. 10:17:38 Checkpoint loguniq 9949, logpos 0x13d5018, timestamp: 0xc4db94= 68 10:23:04 Fuzzy Checkpoint Completed: duration was 26 seconds, 4 buffers= not flushed. 10:23:04 Checkpoint loguniq 9949, logpos 0x13dc080, timestamp: 0xc4e163= 01 10:33:37 Checkpoint Completed: duration was 33 seconds. 10:33:37 Checkpoint loguniq 9949, logpos 0x13dd018, timestamp: 0xc4e92e= 92 10:38:53 Checkpoint Completed: duration was 16 seconds. 10:38:53 Checkpoint loguniq 9949, logpos 0x13e4018, timestamp: 0xc4edef= 80 ***********************************************************************= ******** Forum Note: Use "Reply" to post a response in the discussion forum. =