Checkpoints
in [Oracle]
|
Prev: install Oracle Fusion Middleware 11g meet http 400 error
Next: Force drop of UNDO tablespace (with online segment) in 9.2.0.8?
From: The Magnet on 21 Sep 2009 15:40 Hi, We are on 10gr2. We have 10 redo logs, each 150 megs. We are getting a lot of checkpoint not complete errors. After doing some reading I see that it says not to mix FAST_START_MTTR_TARGET with LOG_CHECKPOINT_INTERVAL. I could keep adding more and more and larger and larger redo logs, but is that really the right answer? I've looked and all apps have decent commits in them......
From: ddf on 21 Sep 2009 15:49 On Sep 21, 2:40 pm, The Magnet <a...(a)unsu.com> wrote: > Hi, > > We are on 10gr2. We have 10 redo logs, each 150 megs. We are getting > a lot of checkpoint not complete errors. After doing some reading I > see that it says not to mix FAST_START_MTTR_TARGET with > LOG_CHECKPOINT_INTERVAL. > > I could keep adding more and more and larger and larger redo logs, but > is that really the right answer? I've looked and all apps have decent > commits in them...... Is log_checkpoint_interval set to a non-zero value? You don't say if it is or if it isn't. And the problem may not be with that parameter but with the size of the redo log groups (they are likely too small and, because of that, you're cycling through the groups at a relatively 'blazing' rate). I would seriously look into increasing the size of those redo logs and aim for two to three log switches per hour. David Fitzjarrell
From: The Magnet on 21 Sep 2009 17:38 On Sep 21, 2:49 pm, ddf <orat...(a)msn.com> wrote: > On Sep 21, 2:40 pm, The Magnet <a...(a)unsu.com> wrote: > > > Hi, > > > We are on 10gr2. We have 10 redo logs, each 150 megs. We are getting > > a lot of checkpoint not complete errors. After doing some reading I > > see that it says not to mix FAST_START_MTTR_TARGET with > > LOG_CHECKPOINT_INTERVAL. > > > I could keep adding more and more and larger and larger redo logs, but > > is that really the right answer? I've looked and all apps have decent > > commits in them...... > > Is log_checkpoint_interval set to a non-zero value? You don't say if > it is or if it isn't. And the problem may not be with that parameter > but with the size of the redo log groups (they are likely too small > and, because of that, you're cycling through the groups at a > relatively 'blazing' rate). I would seriously look into increasing > the size of those redo logs and aim for two to three log switches per > hour. > > David Fitzjarrell Yeah, I forgot to mentions that. The parameter is set to 0. And I should probably find out the frequency of the log switches. I found a query which will do that here: http://www.oraclelog.com/2007/07/17/oracle/determining-log-switch-frequency/ So, let's see if that works.
From: John Hurley on 21 Sep 2009 19:51 On Sep 21, 3:40 pm, The Magnet <a...(a)unsu.com> wrote: snip > We are on 10gr2. We have 10 redo logs, each 150 megs. We are getting > a lot of checkpoint not complete errors. After doing some reading I > see that it says not to mix FAST_START_MTTR_TARGET with > LOG_CHECKPOINT_INTERVAL. > > I could keep adding more and more and larger and larger redo logs, but > is that really the right answer? I've looked and all apps have decent > commits in them...... How many gig of online redo are you generating each day? How much per hour during the busy times? Everyone has different ideas here but most of my OLTP systems have like 20 512 meg files with no real reason not to make them bigger except I like archived logs under a gig.
From: The Magnet on 21 Sep 2009 20:26
On Sep 21, 6:51 pm, John Hurley <johnbhur...(a)sbcglobal.net> wrote: > On Sep 21, 3:40 pm, The Magnet <a...(a)unsu.com> wrote: > > snip > > > We are on 10gr2. We have 10 redo logs, each 150 megs. We are getting > > a lot of checkpoint not complete errors. After doing some reading I > > see that it says not to mix FAST_START_MTTR_TARGET with > > LOG_CHECKPOINT_INTERVAL. > > > I could keep adding more and more and larger and larger redo logs, but > > is that really the right answer? I've looked and all apps have decent > > commits in them...... > > How many gig of online redo are you generating each day? > > How much per hour during the busy times? > > Everyone has different ideas here but most of my OLTP systems have > like 20 512 meg files with no real reason not to make them bigger > except I like archived logs under a gig. Well, I ran a query and it looks like busy times looked like: 8:00am (from 5 - 21 log switches) and 4:00am (from 18 - 35 log switches). Make them larger? Maybe 150 meg is to small? |