Oninit cdi / Archive
Talk to an expert →

long checkpoints in 9.40.fc4

Posted in 2004

Topics: Logging & Checkpoints, Platform-Specific Issues, Versions, Editions & End-of-Life

STEPHEN SCOTT — — source: IIUG Forums & Mailing Lists

We have converted to the following: hpux-11i (64 bit) 4 proc 8 gig ram IDS 9.40.FC4 HP SAN Cooked files This same system ran on a 7.31 database on Solaris (intel)(2 procs) just fine 0-3 second checkpoints. Anybody have any ideas on what I can check/do to fix this. Stephen Scott We have recently had excessive checkpoints. This is a tail -f online.log | grep "Checkpoint Com": 19:42:40 Fuzzy Checkpoint Completed: duration was 214 seconds, 30942 buffers not flushed, 19:51:25 Fuzzy Checkpoint Completed: duration was 1 seconds, 17210 buffers not flushed, 20:13:38 Fuzzy Checkpoint Completed: duration was 1009 seconds, 3259 buffers not flushed, 20:14:09 Fuzzy Checkpoint Completed: duration was 1 seconds, 3263 buffers not flushed, 20:22:44 Fuzzy Checkpoint Completed: duration was 213 seconds, 3264 buffers not flushed, 20:28:10 Fuzzy Checkpoint Completed: duration was 23 seconds, 2204 buffers not flushed, 20:37:52 Fuzzy Checkpoint Completed: duration was 280 seconds, 5945 buffers not flushed, 20:53:09 Fuzzy Checkpoint Completed: duration was 613 seconds, 28922 buffers not flushed, 21:00:11 Fuzzy Checkpoint Completed: duration was 120 seconds, 26170 buffers not flushed, 21:12:38 Fuzzy Checkpoint Completed: duration was 444 seconds, 29681 buffers not flushed, 21:21:06 Checkpoint Completed: duration was 481 seconds. 21:27:18 Checkpoint Completed: duration was 145 seconds. 21:35:11 Checkpoint Completed: duration was 328 seconds. 21:40:52 Fuzzy Checkpoint Completed: duration was 13 seconds, 29257 buffers not flushed, 21:52:49 Checkpoint Completed: duration was 537 seconds. 21:53:28 Checkpoint Completed: duration was 3 seconds.

↪ replying to STEPHEN SCOTT
Gerd Kaluzinski — — source: IIUG Forums & Mailing Lists

On Solaris, it's no big difference between cooked files and rawdevices reagarding I/O-times. On HP it's a big performance loss when you do not user rawdevies. So i think that's your problem as we have seen on some systems. the solutions was alwasy to change to rlvol. gerd STEPHEN SCOTT wrote: >We have converted to the following: >hpux-11i (64 bit) >4 proc >8 gig ram >IDS 9.40.FC4 >HP SAN >Cooked files > >This same system ran on a 7.31 database on Solaris (intel)(2 procs) just fine 0-3 second checkpoints. Anybody have any ideas on what I can check/do to fix this. > >Stephen Scott > >We have recently had excessive checkpoints. > > >This is a tail -f online.log | grep "Checkpoint Com": >19:42:40 Fuzzy Checkpoint Completed: duration was 214 seconds, 30942 buffers not flushed, >19:51:25 Fuzzy Checkpoint Completed: duration was 1 seconds, 17210 buffers not flushed, >20:13:38 Fuzzy Checkpoint Completed: duration was 1009 seconds, 3259 buffers not flushed, >20:14:09 Fuzzy Checkpoint Completed: duration was 1 seconds, 3263 buffers not flushed, >20:22:44 Fuzzy Checkpoint Completed: duration was 213 seconds, 3264 buffers not flushed, >20:28:10 Fuzzy Checkpoint Completed: duration was 23 seconds, 2204 buffers not flushed, >20:37:52 Fuzzy Checkpoint Completed: duration was 280 seconds, 5945 buffers not flushed, >20:53:09 Fuzzy Checkpoint Completed: duration was 613 seconds, 28922 buffers not flushed, >21:00:11 Fuzzy Checkpoint Completed: duration was 120 seconds, 26170 buffers not flushed, >21:12:38 Fuzzy Checkpoint Completed: duration was 444 seconds, 29681 buffers not flushed, >21:21:06 Checkpoint Completed: duration was 481 seconds. >21:27:18 Checkpoint Completed: duration was 145 seconds. >21:35:11 Checkpoint Completed: duration was 328 seconds. >21:40:52 Fuzzy Checkpoint Completed: duration was 13 seconds, 29257 buffers not flushed, >21:52:49 Checkpoint Completed: duration was 537 seconds. >21:53:28 Checkpoint Completed: duration was 3 seconds. > > > > >

↪ replying to STEPHEN SCOTT
Esteban Casuscelli — — source: IIUG Forums & Mailing Lists

Hi, Have you increased the buffer pool? You could try bouncing and configuring the engine to use the standard checkpoint rather than the fuzzy. Have a look at the following onstat ´s outputs during long checkpoints: onstat -R -r 1 | grep dirty onstat -F -r 1 esteban.- At 01:04 a.m. 21/12/2004 -0500, STEPHEN SCOTT wrote: >We have converted to the following: >hpux-11i (64 bit) >4 proc >8 gig ram >IDS 9.40.FC4 >HP SAN >Cooked files > >This same system ran on a 7.31 database on Solaris (intel)(2 procs) just >fine 0-3 second checkpoints. Anybody have any ideas on what I can >check/do to fix this. > >Stephen Scott > >We have recently had excessive checkpoints. > > >This is a tail -f online.log | grep "Checkpoint Com": >19:42:40 Fuzzy Checkpoint Completed: duration was 214 seconds, 30942 >buffers not flushed, >19:51:25 Fuzzy Checkpoint Completed: duration was 1 seconds, 17210 >buffers not flushed, >20:13:38 Fuzzy Checkpoint Completed: duration was 1009 seconds, 3259 >buffers not flushed, >20:14:09 Fuzzy Checkpoint Completed: duration was 1 seconds, 3263 >buffers not flushed, >20:22:44 Fuzzy Checkpoint Completed: duration was 213 seconds, 3264 >buffers not flushed, >20:28:10 Fuzzy Checkpoint Completed: duration was 23 seconds, 2204 >buffers not flushed, >20:37:52 Fuzzy Checkpoint Completed: duration was 280 seconds, 5945 >buffers not flushed, >20:53:09 Fuzzy Checkpoint Completed: duration was 613 seconds, 28922 >buffers not flushed, >21:00:11 Fuzzy Checkpoint Completed: duration was 120 seconds, 26170 >buffers not flushed, >21:12:38 Fuzzy Checkpoint Completed: duration was 444 seconds, 29681 >buffers not flushed, >21:21:06 Checkpoint Completed: duration was 481 seconds. >21:27:18 Checkpoint Completed: duration was 145 seconds. >21:35:11 Checkpoint Completed: duration was 328 seconds. >21:40:52 Fuzzy Checkpoint Completed: duration was 13 seconds, 29257 >buffers not flushed, >21:52:49 Checkpoint Completed: duration was 537 seconds. >21:53:28 Checkpoint Completed: duration was 3 seconds.

Related threads

Glossary