Undocumented checkpoint trigger
Art Kagel saw frequent checkpoints (every 1-5 seconds) during a serial index-build script on 11.70.FC7/RHEL6, labelled with trigger "IPL" in onstat -g ckp, which isn't documented in the Administrator's Reference. Fernando Nunes identified IPL as index page logging: with LOG_INDEX_BUILDS enabled (set because an HDR secondary was planned), each index build forces a checkpoint. Fix: since the parameter is dynamic, leave LOG_INDEX_BUILDS=0 while building indexes and set it to 1 (onmode -wm) only before taking the level-0 backup used to seed secondaries. A doc bug was to be reported.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Folks:
I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
describing checkpoints that are firing every 1-5 seconds during an index
build, with no RTO configured, physical log is 0.1% full, only two logical
logs have filled and been archived via the ALARMPROGRAM immediately.
Checkpoint and current log stamps are never more than on log apart. The
LRU queues are less than 1% dirty (most less than 0.1%).
Does anyone know what's going on?
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
--bcaec54b48d0ff05ca04d85fd578
Index page logging.
Are you running just one create index? AFAIK you'll get one checkpoint for
each index. But if you're creating just one.... Is it on a temporary table?
You don't mention the version and it can make some difference. But there
are people here with more knowledge about this then me...
Regards.
On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com> wrote:
> Folks:
>
> I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> describing checkpoints that are firing every 1-5 seconds during an index
> build, with no RTO configured, physical log is 0.1% full, only two logical
> logs have filled and been archived via the ALARMPROGRAM immediately.
> Checkpoint and current log stamps are never more than on log apart. The
> LRU queues are less than 1% dirty (most less than 0.1%).
>
> Does anyone know what's going on?
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of
> other individuals affiliated with any entity with which I am affiliated nor
> those of the entities themselves.
>
> --bcaec54b48d0ff05ca04d85fd578
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--
Fernando Nunes
Portugal
http://informix-technology.blogspot.com
My email works... but I don't check it frequently...
--047d7b6d934073626904d8600ef8
if you don't know, then no one knows Art! :--)
________________________________
From: Art Kagel <art.kagel@gmail.com>
To: ids@iiug.org
Sent: Wednesday, March 20, 2013 2:00 PM
Subject: Undocumented checkpoint trigger [29810]
Folks:
I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
describing checkpoints that are firing every 1-5 seconds during an index
build, with no RTO configured, physical log is 0.1% full, only two logical
logs have filled and been archived via the ALARMPROGRAM immediately.
Checkpoint and current log stamps are never more than on log apart. The
LRU queues are less than 1% dirty (most less than 0.1%).
Does anyone know what's going on?
Art
Art S. Kagel
Advanced DataTools (http://www.advancedatatools.com/)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
--bcaec54b48d0ff05ca04d85fd578
*******************************************************************************
Forum Note: Use "Reply" to post a response in the discussion forum.
Ooops, forgot <slapping my own wrist>! Version is 11.70.FC7 running on
RHEL6 on Intel x86/64
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
On Wed, Mar 20, 2013 at 3:00 PM, Art Kagel <art.kagel@gmail.com> wrote:
> Folks:
>
> I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> describing checkpoints that are firing every 1-5 seconds during an index
> build, with no RTO configured, physical log is 0.1% full, only two logical
> logs have filled and been archived via the ALARMPROGRAM immediately.
> Checkpoint and current log stamps are never more than on log apart. The
> LRU queues are less than 1% dirty (most less than 0.1%).
>
> Does anyone know what's going on?
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of
> other individuals affiliated with any entity with which I am affiliated nor
> those of the entities themselves.
>
> --bcaec54b48d0ff05ca04d85fd578
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--bcaec554dbc857e42204d8601659
No running a script creating indexes on hundreds of tables, but serially -
one at a time. OK that makes some sense. The instance has no secondary
yet, but it was configured to have one eventually (still in testing) so
index logging is enabled. Thanks Fernando. I'll have to report another
documentation bug now, IPL is not in the lists in the Administrators
Reference.
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
On Wed, Mar 20, 2013 at 3:16 PM, Fernando Nunes <domusonline@gmail.com>wrote:
> Index page logging.
> Are you running just one create index? AFAIK you'll get one checkpoint for
> each index. But if you're creating just one.... Is it on a temporary table?
> You don't mention the version and it can make some difference. But there
> are people here with more knowledge about this then me...
>
> Regards.
>
> On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com> wrote:
>
> > Folks:
> >
> > I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> > describing checkpoints that are firing every 1-5 seconds during an index
> > build, with no RTO configured, physical log is 0.1% full, only two
> logical
> > logs have filled and been archived via the ALARMPROGRAM immediately.
> > Checkpoint and current log stamps are never more than on log apart. The
> > LRU queues are less than 1% dirty (most less than 0.1%).
> >
> > Does anyone know what's going on?
> >
> > Art
> >
> > Art S. Kagel
> > Advanced DataTools (www.advancedatatools.com)
> > Blog: http://informix-myview.blogspot.com/
> >
> > Disclaimer: Please keep in mind that my own opinions are my own opinions
> > and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> > other organization with which I am associated either explicitly,
> > implicitly, or by inference. Neither do those opinions reflect those of
> > other individuals affiliated with any entity with which I am affiliated
> nor
> > those of the entities themselves.
> >
> > --bcaec54b48d0ff05ca04d85fd578
> >
> >
> >
> >
>
>
*******************************************************************************
> > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
>
> --
> Fernando Nunes
> Portugal
>
> http://informix-technology.blogspot.com
> My email works... but I don't check it frequently...
>
> --047d7b6d934073626904d8600ef8
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--485b390f7b20c2e2f204d8602216
I only know stuff 'cause I ask questions... Asked many more long ago and I
still do.
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
On Wed, Mar 20, 2013 at 3:18 PM, Kern Doe <kern_doe@yahoo.com> wrote:
> if you don't know, then no one knows Art! :--)
>
> ________________________________
> From: Art Kagel <art.kagel@gmail.com>
> To: ids@iiug.org
> Sent: Wednesday, March 20, 2013 2:00 PM
> Subject: Undocumented checkpoint trigger [29810]
>
> Folks:
>
> I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> describing checkpoints that are firing every 1-5 seconds during an index
> build, with no RTO configured, physical log is 0.1% full, only two logical
> logs have filled and been archived via the ALARMPROGRAM immediately.
> Checkpoint and current log stamps are never more than on log apart. The
> LRU queues are less than 1% dirty (most less than 0.1%).
>
> Does anyone know what's going on?
>
> Art
>
> Art S. Kagel
> Advanced DataTools (http://www.advancedatatools.com/)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of
> other individuals affiliated with any entity with which I am affiliated nor
> those of the entities themselves.
>
> --bcaec54b48d0ff05ca04d85fd578
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--e89a8f23465924551404d86026ba
I believe before 11.70.FC5 they appeared as "unknown". That was fixed, but
if the docs weren't, than if you don't mind we appreciate.
Regarding LOG_INDEX_BUILDS, just change it to 0 and back to 1 just before
you make the backup that you'll use to setup secondaries.
It's dynamic.. no downtime.
Regards.
On Wed, Mar 20, 2013 at 7:22 PM, Art Kagel <art.kagel@gmail.com> wrote:
> No running a script creating indexes on hundreds of tables, but serially -
> one at a time. OK that makes some sense. The instance has no secondary
> yet, but it was configured to have one eventually (still in testing) so
> index logging is enabled. Thanks Fernando. I'll have to report another
> documentation bug now, IPL is not in the lists in the Administrators
> Reference.
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of
> other individuals affiliated with any entity with which I am affiliated nor
> those of the entities themselves.
>
> On Wed, Mar 20, 2013 at 3:16 PM, Fernando Nunes <domusonline@gmail.com
> >wrote:
>
> > Index page logging.
> > Are you running just one create index? AFAIK you'll get one checkpoint
> for
> > each index. But if you're creating just one.... Is it on a temporary
> table?
> > You don't mention the version and it can make some difference. But there
> > are people here with more knowledge about this then me...
> >
> > Regards.
> >
> > On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com> wrote:
> >
> > > Folks:
> > >
> > > I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> > > describing checkpoints that are firing every 1-5 seconds during an
> index
> > > build, with no RTO configured, physical log is 0.1% full, only two
> > logical
> > > logs have filled and been archived via the ALARMPROGRAM immediately.
> > > Checkpoint and current log stamps are never more than on log apart. The
> > > LRU queues are less than 1% dirty (most less than 0.1%).
> > >
> > > Does anyone know what's going on?
> > >
> > > Art
> > >
> > > Art S. Kagel
> > > Advanced DataTools (www.advancedatatools.com)
> > > Blog: http://informix-myview.blogspot.com/
> > >
> > > Disclaimer: Please keep in mind that my own opinions are my own
> opinions
> > > and do not reflect on my employer, Advanced DataTools, the IIUG, nor
> any
> > > other organization with which I am associated either explicitly,
> > > implicitly, or by inference. Neither do those opinions reflect those of
> > > other individuals affiliated with any entity with which I am affiliated
> > nor
> > > those of the entities themselves.
> > >
> > > --bcaec54b48d0ff05ca04d85fd578
> > >
> > >
> > >
> > >
> >
> >
>
>
*******************************************************************************
> > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > >
> > >
> >
> > --
> > Fernando Nunes
> > Portugal
> >
> > http://informix-technology.blogspot.com
> > My email works... but I don't check it frequently...
> >
> > --047d7b6d934073626904d8600ef8
> >
> >
> >
> >
>
>
*******************************************************************************
> > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
>
> --485b390f7b20c2e2f204d8602216
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--
Fernando Nunes
Portugal
http://informix-technology.blogspot.com
My email works... but I don't check it frequently...
--f46d043d677130cc7b04d862895c
So steps are:
1. onmode -wm LOG_INDEX_BUILDS=0
2. onmode -wm LOG_INDEX_BUILDS=1
3. Archive: ontape -s -L 0
4. Restore on secnodary: ontape -p
5. Setup HDR: onmode -d secondary primserver
6. onmode -d primary secserver
Yes? Seems odd.
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
On Wed, Mar 20, 2013 at 6:13 PM, Fernando Nunes <domusonline@gmail.com>wrote:
> I believe before 11.70.FC5 they appeared as "unknown". That was fixed, but
> if the docs weren't, than if you don't mind we appreciate.
> Regarding LOG_INDEX_BUILDS, just change it to 0 and back to 1 just before
> you make the backup that you'll use to setup secondaries.
> It's dynamic.. no downtime.
> Regards.
>
> On Wed, Mar 20, 2013 at 7:22 PM, Art Kagel <art.kagel@gmail.com> wrote:
>
> > No running a script creating indexes on hundreds of tables, but serially
> -
> > one at a time. OK that makes some sense. The instance has no secondary
> > yet, but it was configured to have one eventually (still in testing) so
> > index logging is enabled. Thanks Fernando. I'll have to report another
> > documentation bug now, IPL is not in the lists in the Administrators
> > Reference.
> >
> > Art
> >
> > Art S. Kagel
> > Advanced DataTools (www.advancedatatools.com)
> > Blog: http://informix-myview.blogspot.com/
> >
> > Disclaimer: Please keep in mind that my own opinions are my own opinions
> > and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> > other organization with which I am associated either explicitly,
> > implicitly, or by inference. Neither do those opinions reflect those of
> > other individuals affiliated with any entity with which I am affiliated
> nor
> > those of the entities themselves.
> >
> > On Wed, Mar 20, 2013 at 3:16 PM, Fernando Nunes <domusonline@gmail.com
> > >wrote:
> >
> > > Index page logging.
> > > Are you running just one create index? AFAIK you'll get one checkpoint
> > for
> > > each index. But if you're creating just one.... Is it on a temporary
> > table?
> > > You don't mention the version and it can make some difference. But
> there
> > > are people here with more knowledge about this then me...
> > >
> > > Regards.
> > >
> > > On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com>
> wrote:
> > >
> > > > Folks:
> > > >
> > > > I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> > > > describing checkpoints that are firing every 1-5 seconds during an
> > index
> > > > build, with no RTO configured, physical log is 0.1% full, only two
> > > logical
> > > > logs have filled and been archived via the ALARMPROGRAM immediately.
> > > > Checkpoint and current log stamps are never more than on log apart.
> The
> > > > LRU queues are less than 1% dirty (most less than 0.1%).
> > > >
> > > > Does anyone know what's going on?
> > > >
> > > > Art
> > > >
> > > > Art S. Kagel
> > > > Advanced DataTools (www.advancedatatools.com)
> > > > Blog: http://informix-myview.blogspot.com/
> > > >
> > > > Disclaimer: Please keep in mind that my own opinions are my own
> > opinions
> > > > and do not reflect on my employer, Advanced DataTools, the IIUG, nor
> > any
> > > > other organization with which I am associated either explicitly,
> > > > implicitly, or by inference. Neither do those opinions reflect those
> of
> > > > other individuals affiliated with any entity with which I am
> affiliated
> > > nor
> > > > those of the entities themselves.
> > > >
> > > > --bcaec54b48d0ff05ca04d85fd578
> > > >
> > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>
*******************************************************************************
> > > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > > >
> > > >
> > >
> > > --
> > > Fernando Nunes
> > > Portugal
> > >
> > > http://informix-technology.blogspot.com
> > > My email works... but I don't check it frequently...
> > >
> > > --047d7b6d934073626904d8600ef8
> > >
> > >
> > >
> > >
> >
> >
>
>
*******************************************************************************
> > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > >
> > >
> >
> > --485b390f7b20c2e2f204d8602216
> >
> >
> >
> >
>
>
*******************************************************************************
> > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
>
> --
> Fernando Nunes
> Portugal
>
> http://informix-technology.blogspot.com
> My email works... but I don't check it frequently...
>
> --f46d043d677130cc7b04d862895c
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--f46d042d051cfdac4604d8631701
It's funny how after some years of writing a foreign language without
having proper training we tend to think that we know how to do it...
Sometimes we (I) mess up.
What you describe is what I've written and yes, it looks odd because I
choose my words poorly or missed some punctuation.
What I intended to write was:
You don't need to have LOG_INDEX_BUILDS=1 just because you're thinking
about setting up secondary servers in the future. You just need to make
sure that before you create the backup you have it set to one. Before that
you can have the engine with LOG_INDEX_BUILDS=0 and that will not cause
those kind of checkpoints while you run the script that creates the indexes.
So...
1- onmode -wm LOG_INDEX_BUILDS=0
[... work with the engine as long as you like in standard mode ...]
[... once you decide to create the secondary server(s)....]
2- onmode -wm LOG_INDEX_BUILDS=1
3- ontape -s -L 0
and etc.
Regards
On Wed, Mar 20, 2013 at 10:53 PM, Art Kagel <art.kagel@gmail.com> wrote:
> So steps are:
>
> 1. onmode -wm LOG_INDEX_BUILDS=0
>
> 2. onmode -wm LOG_INDEX_BUILDS=1
>
> 3. Archive: ontape -s -L 0
>
> 4. Restore on secnodary: ontape -p
>
> 5. Setup HDR: onmode -d secondary primserver
>
> 6. onmode -d primary secserver
>
> Yes? Seems odd.
>
> Art
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of
> other individuals affiliated with any entity with which I am affiliated nor
> those of the entities themselves.
>
> On Wed, Mar 20, 2013 at 6:13 PM, Fernando Nunes <domusonline@gmail.com
> >wrote:
>
> > I believe before 11.70.FC5 they appeared as "unknown". That was fixed,
> but
> > if the docs weren't, than if you don't mind we appreciate.
> > Regarding LOG_INDEX_BUILDS, just change it to 0 and back to 1 just before
> > you make the backup that you'll use to setup secondaries.
> > It's dynamic.. no downtime.
> > Regards.
> >
> > On Wed, Mar 20, 2013 at 7:22 PM, Art Kagel <art.kagel@gmail.com> wrote:
> >
> > > No running a script creating indexes on hundreds of tables, but
> serially
> > -
> > > one at a time. OK that makes some sense. The instance has no secondary
> > > yet, but it was configured to have one eventually (still in testing) so
> > > index logging is enabled. Thanks Fernando. I'll have to report another
> > > documentation bug now, IPL is not in the lists in the Administrators
> > > Reference.
> > >
> > > Art
> > >
> > > Art S. Kagel
> > > Advanced DataTools (www.advancedatatools.com)
> > > Blog: http://informix-myview.blogspot.com/
> > >
> > > Disclaimer: Please keep in mind that my own opinions are my own
> opinions
> > > and do not reflect on my employer, Advanced DataTools, the IIUG, nor
> any
> > > other organization with which I am associated either explicitly,
> > > implicitly, or by inference. Neither do those opinions reflect those of
> > > other individuals affiliated with any entity with which I am affiliated
> > nor
> > > those of the entities themselves.
> > >
> > > On Wed, Mar 20, 2013 at 3:16 PM, Fernando Nunes <domusonline@gmail.com
> > > >wrote:
> > >
> > > > Index page logging.
> > > > Are you running just one create index? AFAIK you'll get one
> checkpoint
> > > for
> > > > each index. But if you're creating just one.... Is it on a temporary
> > > table?
> > > > You don't mention the version and it can make some difference. But
> > there
> > > > are people here with more knowledge about this then me...
> > > >
> > > > Regards.
> > > >
> > > > On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com>
> > wrote:
> > > >
> > > > > Folks:
> > > > >
> > > > > I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> > > > > describing checkpoints that are firing every 1-5 seconds during an
> > > index
> > > > > build, with no RTO configured, physical log is 0.1% full, only two
> > > > logical
> > > > > logs have filled and been archived via the ALARMPROGRAM
> immediately.
> > > > > Checkpoint and current log stamps are never more than on log apart.
> > The
> > > > > LRU queues are less than 1% dirty (most less than 0.1%).
> > > > >
> > > > > Does anyone know what's going on?
> > > > >
> > > > > Art
> > > > >
> > > > > Art S. Kagel
> > > > > Advanced DataTools (www.advancedatatools.com)
> > > > > Blog: http://informix-myview.blogspot.com/
> > > > >
> > > > > Disclaimer: Please keep in mind that my own opinions are my own
> > > opinions
> > > > > and do not reflect on my employer, Advanced DataTools, the IIUG,
> nor
> > > any
> > > > > other organization with which I am associated either explicitly,
> > > > > implicitly, or by inference. Neither do those opinions reflect
> those
> > of
> > > > > other individuals affiliated with any entity with which I am
> > affiliated
> > > > nor
> > > > > those of the entities themselves.
> > > > >
> > > > > --bcaec54b48d0ff05ca04d85fd578
> > > > >
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>
*******************************************************************************
> > > > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > > > >
> > > > >
> > > >
> > > > --
> > > > Fernando Nunes
> > > > Portugal
> > > >
> > > > http://informix-technology.blogspot.com
> > > > My email works... but I don't check it frequently...
> > > >
> > > > --047d7b6d934073626904d8600ef8
> > > >
> > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>
*******************************************************************************
> > > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > > >
> > > >
> > >
> > > --485b390f7b20c2e2f204d8602216
> > >
> > >
> > >
> > >
> >
> >
>
>
*******************************************************************************
> > > Forum Note: Use "Reply" to post a response in the discussion forum.
> > >
> > >
> >
> > --
> > Fernando Nunes
> > Portugal
> >
> > http://informix-technology.blogspot.com
> > My email works... but I don't check it frequently...
> >
> > --f46d043d677130cc7b04d862895c
> >
> >
> >
> >
>
>
*******************************************************************************
> > Forum Note: Use "Reply" to post a response in the discussion forum.
> >
> >
>
> --f46d042d051cfdac4604d8631701
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--
Fern
Ahh, I suspected it was just a disconnect but wanted to check. The
configuration is already set up for HDR because the ONCONFIG was cloned
from the production server (this will be the QA server) and that server is
already logged.
Executing a checkpoint everytime an index build completes seems excessive
to me. Leaves me wondering why someone found that necessary. Retorical
question.
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
On Wed, Mar 20, 2013 at 7:22 PM, Fernando Nunes <domusonline@gmail.com>wrote:
> It's funny how after some years of writing a foreign language without
> having proper training we tend to think that we know how to do it...
> Sometimes we (I) mess up.
> What you describe is what I've written and yes, it looks odd because I
> choose my words poorly or missed some punctuation.
> What I intended to write was:
>
> You don't need to have LOG_INDEX_BUILDS=1 just because you're thinking
> about setting up secondary servers in the future. You just need to make
> sure that before you create the backup you have it set to one. Before that
> you can have the engine with LOG_INDEX_BUILDS=0 and that will not cause
> those kind of checkpoints while you run the script that creates the
> indexes.
>
> So...
> 1- onmode -wm LOG_INDEX_BUILDS=0
>
> [... work with the engine as long as you like in standard mode ...]
> [... once you decide to create the secondary server(s)....]
> 2- onmode -wm LOG_INDEX_BUILDS=1
> 3- ontape -s -L 0
> and etc.
>
> Regards
>
> On Wed, Mar 20, 2013 at 10:53 PM, Art Kagel <art.kagel@gmail.com> wrote:
>
> > So steps are:
> >
> > 1. onmode -wm LOG_INDEX_BUILDS=0
> >
> > 2. onmode -wm LOG_INDEX_BUILDS=1
> >
> > 3. Archive: ontape -s -L 0
> >
> > 4. Restore on secnodary: ontape -p
> >
> > 5. Setup HDR: onmode -d secondary primserver
> >
> > 6. onmode -d primary secserver
> >
> > Yes? Seems odd.
> >
> > Art
> > Art S. Kagel
> > Advanced DataTools (www.advancedatatools.com)
> > Blog: http://informix-myview.blogspot.com/
> >
> > Disclaimer: Please keep in mind that my own opinions are my own opinions
> > and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> > other organization with which I am associated either explicitly,
> > implicitly, or by inference. Neither do those opinions reflect those of
> > other individuals affiliated with any entity with which I am affiliated
> nor
> > those of the entities themselves.
> >
> > On Wed, Mar 20, 2013 at 6:13 PM, Fernando Nunes <domusonline@gmail.com
> > >wrote:
> >
> > > I believe before 11.70.FC5 they appeared as "unknown". That was fixed,
> > but
> > > if the docs weren't, than if you don't mind we appreciate.
> > > Regarding LOG_INDEX_BUILDS, just change it to 0 and back to 1 just
> before
> > > you make the backup that you'll use to setup secondaries.
> > > It's dynamic.. no downtime.
> > > Regards.
> > >
> > > On Wed, Mar 20, 2013 at 7:22 PM, Art Kagel <art.kagel@gmail.com>
> wrote:
> > >
> > > > No running a script creating indexes on hundreds of tables, but
> > serially
> > > -
> > > > one at a time. OK that makes some sense. The instance has no
> secondary
> > > > yet, but it was configured to have one eventually (still in testing)
> so
> > > > index logging is enabled. Thanks Fernando. I'll have to report
> another
> > > > documentation bug now, IPL is not in the lists in the Administrators
> > > > Reference.
> > > >
> > > > Art
> > > >
> > > > Art S. Kagel
> > > > Advanced DataTools (www.advancedatatools.com)
> > > > Blog: http://informix-myview.blogspot.com/
> > > >
> > > > Disclaimer: Please keep in mind that my own opinions are my own
> > opinions
> > > > and do not reflect on my employer, Advanced DataTools, the IIUG, nor
> > any
> > > > other organization with which I am associated either explicitly,
> > > > implicitly, or by inference. Neither do those opinions reflect those
> of
> > > > other individuals affiliated with any entity with which I am
> affiliated
> > > nor
> > > > those of the entities themselves.
> > > >
> > > > On Wed, Mar 20, 2013 at 3:16 PM, Fernando Nunes <
> domusonline@gmail.com
> > > > >wrote:
> > > >
> > > > > Index page logging.
> > > > > Are you running just one create index? AFAIK you'll get one
> > checkpoint
> > > > for
> > > > > each index. But if you're creating just one.... Is it on a
> temporary
> > > > table?
> > > > > You don't mention the version and it can make some difference. But
> > > there
> > > > > are people here with more knowledge about this then me...
> > > > >
> > > > > Regards.
> > > > >
> > > > > On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com>
> > > wrote:
> > > > >
> > > > > > Folks:
> > > > > >
> > > > > > I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> > > > > > describing checkpoints that are firing every 1-5 seconds during
> an
> > > > index
> > > > > > build, with no RTO configured, physical log is 0.1% full, only
> two
> > > > > logical
> > > > > > logs have filled and been archived via the ALARMPROGRAM
> > immediately.
> > > > > > Checkpoint and current log stamps are never more than on log
> apart.
> > > The
> > > > > > LRU queues are less than 1% dirty (most less than 0.1%).
> > > > > >
> > > > > > Does anyone know what's going on?
> > > > > >
> > > > > > Art
> > > > > >
> > > > > > Art S. Kagel
> > > > > > Advanced DataTools (www.advancedatatools.com)
> > > > > > Blog: http://informix-myview.blogspot.com/
> > > > > >
> > > > > > Disclaimer: Please keep in mind that my own opinions are my own
> > > > opinions
> > > > > > and do not reflect on my employer, Advanced DataTools, the IIUG,
> > nor
> > > > any
> > > > > > other organization with which I am associated either explicitly,
> > > > > > implicitly, or by inference. Neither do those opinions reflect
> > those
> > > of
> > > > > > other individuals affiliated with any entity with which I am
> > > affiliated
> > > > > nor
> > > > > > those of the entities themselves.
> > > > > >
> > > > > > --bcaec54b48d0ff05ca04d85fd578
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>
*******************************************************************************
> > > > > > Forum Note: Use "Reply" to post a response in the discussion
> forum.
> > > > > >
> > > > > >
> > > > >
> > > > > --
> > > > > Fernando Nunes
> > > > > Portugal
> > > > >
> > > > > http://infor
It has to do with what happens on the secondary servers when they receive
and process the logs. I didn't check the details, but it was done to avoid
bigger problems.
It only happens with LOG_INDEX_BUILDS active.
And you know who can give you all the details :)
Regards.
On Thu, Mar 21, 2013 at 12:46 AM, Art Kagel <art.kagel@gmail.com> wrote:
> Ahh, I suspected it was just a disconnect but wanted to check. The
> configuration is already set up for HDR because the ONCONFIG was cloned
> from the production server (this will be the QA server) and that server is
> already logged.
>
> Executing a checkpoint everytime an index build completes seems excessive
> to me. Leaves me wondering why someone found that necessary. Retorical
> question.
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinions
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of
> other individuals affiliated with any entity with which I am affiliated nor
> those of the entities themselves.
>
> On Wed, Mar 20, 2013 at 7:22 PM, Fernando Nunes <domusonline@gmail.com
> >wrote:
>
> > It's funny how after some years of writing a foreign language without
> > having proper training we tend to think that we know how to do it...
> > Sometimes we (I) mess up.
> > What you describe is what I've written and yes, it looks odd because I
> > choose my words poorly or missed some punctuation.
> > What I intended to write was:
> >
> > You don't need to have LOG_INDEX_BUILDS=1 just because you're thinking
> > about setting up secondary servers in the future. You just need to make
> > sure that before you create the backup you have it set to one. Before
> that
> > you can have the engine with LOG_INDEX_BUILDS=0 and that will not cause
> > those kind of checkpoints while you run the script that creates the
> > indexes.
> >
> > So...
> > 1- onmode -wm LOG_INDEX_BUILDS=0
> >
> > [... work with the engine as long as you like in standard mode ...]
> > [... once you decide to create the secondary server(s)....]
> > 2- onmode -wm LOG_INDEX_BUILDS=1
> > 3- ontape -s -L 0
> > and etc.
> >
> > Regards
> >
> > On Wed, Mar 20, 2013 at 10:53 PM, Art Kagel <art.kagel@gmail.com> wrote:
> >
> > > So steps are:
> > >
> > > 1. onmode -wm LOG_INDEX_BUILDS=0
> > >
> > > 2. onmode -wm LOG_INDEX_BUILDS=1
> > >
> > > 3. Archive: ontape -s -L 0
> > >
> > > 4. Restore on secnodary: ontape -p
> > >
> > > 5. Setup HDR: onmode -d secondary primserver
> > >
> > > 6. onmode -d primary secserver
> > >
> > > Yes? Seems odd.
> > >
> > > Art
> > > Art S. Kagel
> > > Advanced DataTools (www.advancedatatools.com)
> > > Blog: http://informix-myview.blogspot.com/
> > >
> > > Disclaimer: Please keep in mind that my own opinions are my own
> opinions
> > > and do not reflect on my employer, Advanced DataTools, the IIUG, nor
> any
> > > other organization with which I am associated either explicitly,
> > > implicitly, or by inference. Neither do those opinions reflect those of
> > > other individuals affiliated with any entity with which I am affiliated
> > nor
> > > those of the entities themselves.
> > >
> > > On Wed, Mar 20, 2013 at 6:13 PM, Fernando Nunes <domusonline@gmail.com
> > > >wrote:
> > >
> > > > I believe before 11.70.FC5 they appeared as "unknown". That was
> fixed,
> > > but
> > > > if the docs weren't, than if you don't mind we appreciate.
> > > > Regarding LOG_INDEX_BUILDS, just change it to 0 and back to 1 just
> > before
> > > > you make the backup that you'll use to setup secondaries.
> > > > It's dynamic.. no downtime.
> > > > Regards.
> > > >
> > > > On Wed, Mar 20, 2013 at 7:22 PM, Art Kagel <art.kagel@gmail.com>
> > wrote:
> > > >
> > > > > No running a script creating indexes on hundreds of tables, but
> > > serially
> > > > -
> > > > > one at a time. OK that makes some sense. The instance has no
> > secondary
> > > > > yet, but it was configured to have one eventually (still in
> testing)
> > so
> > > > > index logging is enabled. Thanks Fernando. I'll have to report
> > another
> > > > > documentation bug now, IPL is not in the lists in the
> Administrators
> > > > > Reference.
> > > > >
> > > > > Art
> > > > >
> > > > > Art S. Kagel
> > > > > Advanced DataTools (www.advancedatatools.com)
> > > > > Blog: http://informix-myview.blogspot.com/
> > > > >
> > > > > Disclaimer: Please keep in mind that my own opinions are my own
> > > opinions
> > > > > and do not reflect on my employer, Advanced DataTools, the IIUG,
> nor
> > > any
> > > > > other organization with which I am associated either explicitly,
> > > > > implicitly, or by inference. Neither do those opinions reflect
> those
> > of
> > > > > other individuals affiliated with any entity with which I am
> > affiliated
> > > > nor
> > > > > those of the entities themselves.
> > > > >
> > > > > On Wed, Mar 20, 2013 at 3:16 PM, Fernando Nunes <
> > domusonline@gmail.com
> > > > > >wrote:
> > > > >
> > > > > > Index page logging.
> > > > > > Are you running just one create index? AFAIK you'll get one
> > > checkpoint
> > > > > for
> > > > > > each index. But if you're creating just one.... Is it on a
> > temporary
> > > > > table?
> > > > > > You don't mention the version and it can make some difference.
> But
> > > > there
> > > > > > are people here with more knowledge about this then me...
> > > > > >
> > > > > > Regards.
> > > > > >
> > > > > > On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com>
> > > > wrote:
> > > > > >
> > > > > > > Folks:
> > > > > > >
> > > > > > > I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp
> report
> > > > > > > describing checkpoints that are firing every 1-5 seconds during
> > an
> > > > > index
> > > > > > > build, with no RTO configured, physical log is 0.1% full, only
> > two
> > > > > > logical
> > > > > > > logs have filled and been archived via the ALARMPROGRAM
> > > immediately.
> > > > > > > Checkpoint and current log stamps are never more than on log
> > apart.
> > > > The
> > > > > > > LRU queues are less than 1% dirty (most less than 0.1%).
> > > > > > >
> > > > > > > Does anyone know what's going on?
> > > > > > >
> > > > > > > Art
> > > > > > >
> > > > > > > Art S. Kagel
> > > > > > > Advanced DataTools (www.advancedatatools.com)
> > > > > > > Blog: http://informix-myview.blogspot.com/
> > > > > > >
> > > > > > > Disclaimer: Please keep in mind that my own opinions are my own
> > > > > opinions
> > > > > > > and do not reflect on my employer, Advanced DataTools, the
> IIUG,
> > > nor
> > > > > any
> > > > > > > other organization with which I am associated either
> explicitly,
> > > > > > > implicitly, or by inference
It's to keep the secondary buffer pool from getting too dirty when the
index is being transferred to the secondary.
Since the index is built on the primary and then transferer to the
secondary, then the primary has already encountered the cost of maintai=
ning
the buffer pool while the index was being built. That is not the case =
on
the secondary. So there are a couple of things that we do to prevent t=
he
index transfer from creating a back flow which would potentially impact=
the
primary.
We take two checkpoints at the start of the transfer. This is necessar=
y to
prevent us from having to PLOG the pages on the secondary which are abo=
ut
to become index pages. This would be the case otherwise if there had b=
een
any deleted partitions ....
Then we create non-blocking checkpoints periodically during the index
transfer. This serves two purposes. 1) It reduces the work that has t=
o be
done in the event of a system failure and 2) it keeps the secondary cle=
an
during the index transfer in the most efficient means.
We do the checkpoints frequently to avoid excessive impact on the prima=
ry
during the transfer - and also by doing them fairly frequently. We
actually have to force a dirty page with these IPL checkpoints because =
in
most cases there is a probability that there are otherwise no dirty pag=
es
on the primary and that would cause the primary to treat the checkpoint=
request as a NOP.
Again don't forget that during the index transfer, the primary is readi=
ng
the index that it previously created, but the secondary is writing. An=
d
that creates an imbalance in the generation of dirty pages which the in=
dex
transfer is underway.
From: "Art Kagel" <art.kagel@gmail.com>
To: ids@iiug.org,
Date: 03/20/2013 02:01 PM
Subject: Undocumented checkpoint trigger [29810]
Sent by: ids-bounces@iiug.org
Folks:
I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
describing checkpoints that are firing every 1-5 seconds during an inde=
x
build, with no RTO configured, physical log is 0.1% full, only two logi=
cal
logs have filled and been archived via the ALARMPROGRAM immediately.
Checkpoint and current log stamps are never more than on log apart. The=
LRU queues are less than 1% dirty (most less than 0.1%).
Does anyone know what's going on?
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinion=
s
and do not reflect on my employer, Advanced DataTools, the IIUG, nor an=
y
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of=
other individuals affiliated with any entity with which I am affiliated=
nor
those of the entities themselves.
--bcaec54b48d0ff05ca04d85fd578
***********************************************************************=
********
Forum Note: Use "Reply" to post a response in the discussion forum.
=
Don't you think you could explain that better :-)
Cheers
Paul
Paul Watson
Oninit www.oninit.com
+1 913 387 7529
On Mar 20, 2013, at 21:54, "Madison Pruet" <mpruet@us.ibm.com> wrote:
> It's to keep the secondary buffer pool from getting too dirty when the
> index is being transferred to the secondary.
>
> Since the index is built on the primary and then transferer to the
> secondary, then the primary has already encountered the cost of maintai=
> ning
> the buffer pool while the index was being built. That is not the case =
> on
> the secondary. So there are a couple of things that we do to prevent t=
> he
> index transfer from creating a back flow which would potentially impact=
> the
> primary.
>
> We take two checkpoints at the start of the transfer. This is necessar=
> y to
> prevent us from having to PLOG the pages on the secondary which are abo=
> ut
> to become index pages. This would be the case otherwise if there had b=
> een
> any deleted partitions ....
>
> Then we create non-blocking checkpoints periodically during the index
> transfer. This serves two purposes. 1) It reduces the work that has t=
> o be
> done in the event of a system failure and 2) it keeps the secondary cle=
> an
> during the index transfer in the most efficient means.
>
> We do the checkpoints frequently to avoid excessive impact on the prima=
> ry
> during the transfer - and also by doing them fairly frequently. We
> actually have to force a dirty page with these IPL checkpoints because =
> in
> most cases there is a probability that there are otherwise no dirty pag=
> es
> on the primary and that would cause the primary to treat the checkpoint=
>
> request as a NOP.
>
> Again don't forget that during the index transfer, the primary is readi=
> ng
> the index that it previously created, but the secondary is writing. An=
> d
> that creates an imbalance in the generation of dirty pages which the in=
> dex
> transfer is underway.
>
> From: "Art Kagel" <art.kagel@gmail.com>
> To: ids@iiug.org,
> Date: 03/20/2013 02:01 PM
> Subject: Undocumented checkpoint trigger [29810]
> Sent by: ids-bounces@iiug.org
>
> Folks:
>
> I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> describing checkpoints that are firing every 1-5 seconds during an inde=
> x
> build, with no RTO configured, physical log is 0.1% full, only two logi=
> cal
> logs have filled and been archived via the ALARMPROGRAM immediately.
> Checkpoint and current log stamps are never more than on log apart. The=
>
> LRU queues are less than 1% dirty (most less than 0.1%).
>
> Does anyone know what's going on?
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinion=
> s
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor an=
> y
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of=
>
> other individuals affiliated with any entity with which I am affiliated=
> nor
>
> those of the entities themselves.
>
> --bcaec54b48d0ff05ca04d85fd578
>
> ***********************************************************************=
> ********
>
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
> =
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
I'm not sure that really makes a difference. We ran into quite a few
problems in the past with both IPL index transfers as well as the older=
HDR
index transfers. The main problems all came down the to the fundamenta=
l
imbalance in dirty page generation on the primary/secondary during the
index transfer. At that point in time, the primary would be basically
doing IO reads compared to all of the writes being done on the secondar=
y.
Of course, during the primary would have been rather busy maintaining a=
clean buffer pool during the index creation, but the secondary is faced=
with that task during the index transfer. And as I mentioned in my earl=
ier
email - by doing them frequently, we are minimizing the impact on the
primary while keeping the secondary buffer pool clean.
From: "Fernando Nunes" <domusonline@gmail.com>
To: ids@iiug.org,
Date: 03/20/2013 05:15 PM
Subject: Re: Undocumented checkpoint trigger [29822]
Sent by: ids-bounces@iiug.org
I believe before 11.70.FC5 they appeared as "unknown". That was fixed, =
but
if the docs weren't, than if you don't mind we appreciate.
Regarding LOG_INDEX_BUILDS, just change it to 0 and back to 1 just befo=
re
you make the backup that you'll use to setup secondaries.
It's dynamic.. no downtime.
Regards.
On Wed, Mar 20, 2013 at 7:22 PM, Art Kagel <art.kagel@gmail.com> wrote:=
> No running a script creating indexes on hundreds of tables, but seria=
lly
-
> one at a time. OK that makes some sense. The instance has no secondar=
y
> yet, but it was configured to have one eventually (still in testing) =
so
> index logging is enabled. Thanks Fernando. I'll have to report anothe=
r
> documentation bug now, IPL is not in the lists in the Administrators
> Reference.
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opini=
ons
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor =
any
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those =
of
> other individuals affiliated with any entity with which I am affiliat=
ed
nor
> those of the entities themselves.
>
> On Wed, Mar 20, 2013 at 3:16 PM, Fernando Nunes <domusonline@gmail.co=
m
> >wrote:
>
> > Index page logging.
> > Are you running just one create index? AFAIK you'll get one checkpo=
int
> for
> > each index. But if you're creating just one.... Is it on a temporar=
y
> table?
> > You don't mention the version and it can make some difference. But
there
> > are people here with more knowledge about this then me...
> >
> > Regards.
> >
> > On Wed, Mar 20, 2013 at 7:00 PM, Art Kagel <art.kagel@gmail.com> wr=
ote:
> >
> > > Folks:
> > >
> > > I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report=
> > > describing checkpoints that are firing every 1-5 seconds during a=
n
> index
> > > build, with no RTO configured, physical log is 0.1% full, only tw=
o
> > logical
> > > logs have filled and been archived via the ALARMPROGRAM immediate=
ly.
> > > Checkpoint and current log stamps are never more than on log apar=
t.
The
> > > LRU queues are less than 1% dirty (most less than 0.1%).
> > >
> > > Does anyone know what's going on?
> > >
> > > Art
> > >
> > > Art S. Kagel
> > > Advanced DataTools (www.advancedatatools.com)
> > > Blog: http://informix-myview.blogspot.com/
> > >
> > > Disclaimer: Please keep in mind that my own opinions are my own
> opinions
> > > and do not reflect on my employer, Advanced DataTools, the IIUG, =
nor
> any
> > > other organization with which I am associated either explicitly,
> > > implicitly, or by inference. Neither do those opinions reflect th=
ose
of
> > > other individuals affiliated with any entity with which I am
affiliated
> > nor
> > > those of the entities themselves.
> > >
> > > --bcaec54b48d0ff05ca04d85fd578
> > >
> > >
> > >
> > >
> >
> >
>
>
***********************************************************************=
********
> > > Forum Note: Use "Reply" to post a response in the discussion foru=
m.
> > >
> > >
> >
> > --
> > Fernando Nunes
> > Portugal
> >
> > http://informix-technology.blogspot.com
> > My email works... but I don't check it frequently...
> >
> > --047d7b6d934073626904d8600ef8
> >
> >
> >
> >
>
>
***********************************************************************=
********
> > Forum Note: Use "Reply" to post a response in the discussion forum.=
> >
> >
>
> --485b390f7b20c2e2f204d8602216
>
>
>
>
***********************************************************************=
********
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--
Fernando Nunes
Portugal
http://informix-technology.blogspot.com
My email works... but I don't check it frequently...
--f46d043d677130cc7b04d862895c
***********************************************************************=
********
Forum Note: Use "Reply" to post a response in the discussion forum.
=
Thanks Madison. I'll have to chew on that for a bit to digest it, but it
makes more sense now.
Art
Art S. Kagel
Advanced DataTools (www.advancedatatools.com)
Blog: http://informix-myview.blogspot.com/
Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on my employer, Advanced DataTools, the IIUG, nor any
other organization with which I am associated either explicitly,
implicitly, or by inference. Neither do those opinions reflect those of
other individuals affiliated with any entity with which I am affiliated nor
those of the entities themselves.
On Wed, Mar 20, 2013 at 10:54 PM, Madison Pruet <mpruet@us.ibm.com> wrote:
> It's to keep the secondary buffer pool from getting too dirty when the
> index is being transferred to the secondary.
>
> Since the index is built on the primary and then transferer to the
> secondary, then the primary has already encountered the cost of maintai=
> ning
> the buffer pool while the index was being built. That is not the case =
> on
> the secondary. So there are a couple of things that we do to prevent t=
> he
> index transfer from creating a back flow which would potentially impact=
> the
> primary.
>
> We take two checkpoints at the start of the transfer. This is necessar=
> y to
> prevent us from having to PLOG the pages on the secondary which are abo=
> ut
> to become index pages. This would be the case otherwise if there had b=
> een
> any deleted partitions ....
>
> Then we create non-blocking checkpoints periodically during the index
> transfer. This serves two purposes. 1) It reduces the work that has t=
> o be
> done in the event of a system failure and 2) it keeps the secondary cle=
> an
> during the index transfer in the most efficient means.
>
> We do the checkpoints frequently to avoid excessive impact on the prima=
> ry
> during the transfer - and also by doing them fairly frequently. We
> actually have to force a dirty page with these IPL checkpoints because =
> in
> most cases there is a probability that there are otherwise no dirty pag=
> es
> on the primary and that would cause the primary to treat the checkpoint=
>
> request as a NOP.
>
> Again don't forget that during the index transfer, the primary is readi=
> ng
> the index that it previously created, but the secondary is writing. An=
> d
> that creates an imbalance in the generation of dirty pages which the in=
> dex
> transfer is underway.
>
> From: "Art Kagel" <art.kagel@gmail.com>
> To: ids@iiug.org,
> Date: 03/20/2013 02:01 PM
> Subject: Undocumented checkpoint trigger [29810]
> Sent by: ids-bounces@iiug.org
>
> Folks:
>
> I'm seeing a checkpoint trigger "IPL" in the onstat -g ckp report
> describing checkpoints that are firing every 1-5 seconds during an inde=
> x
> build, with no RTO configured, physical log is 0.1% full, only two logi=
> cal
> logs have filled and been archived via the ALARMPROGRAM immediately.
> Checkpoint and current log stamps are never more than on log apart. The=
>
> LRU queues are less than 1% dirty (most less than 0.1%).
>
> Does anyone know what's going on?
>
> Art
>
> Art S. Kagel
> Advanced DataTools (www.advancedatatools.com)
> Blog: http://informix-myview.blogspot.com/
>
> Disclaimer: Please keep in mind that my own opinions are my own opinion=
> s
> and do not reflect on my employer, Advanced DataTools, the IIUG, nor an=
> y
> other organization with which I am associated either explicitly,
> implicitly, or by inference. Neither do those opinions reflect those of=
>
> other individuals affiliated with any entity with which I am affiliated=
> nor
>
> those of the entities themselves.
>
> --bcaec54b48d0ff05ca04d85fd578
>
> ***********************************************************************=
> ********
>
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
> =
>
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
--bcaec554d232ac3f3404d86c6247
Related threads
- Posting from the Informix-list
- Migrating from IDS 9.40.UC6 to 11.50.UC3
- Ip for a network session
- questions onstat -g