# of page steal cleaner will trigger page threshold?

# of page steal cleaner will trigger page threshold?

Post by Fan Ruo Xi » Wed, 14 Aug 2002 10:55:44



Hello,
Is it db2 the same thing like informix in this way when a couple of
dirty page steal cleaner triggers happen, db2 will trigger dirty page
threshold cleaner trigger, even the dirty pages in dirty queue have
touched the page threshold???

From my test, it looks like db2 does not use this way. But I am not
sure. If db2 use is like this, how many number of dirty page steal
cleaner trggers will cause the dirty page threshold cleaner trigger???

Best Regards,
Fan Ruo Xin

 
 
 

# of page steal cleaner will trigger page threshold?

Post by <aam.. » Mon, 19 Aug 2002 05:40:57


No.

Dirty steal triggers and dirty threshold triggers are completely
different.  Threshold will only trigger at the change page threshold
that you have setup in the db cfg.

A dirty steal trigger will occur when a db2 agent victimizes
a page which happens to be dirty.


Quote:> Hello,
> Is it db2 the same thing like informix in this way when a couple of
> dirty page steal cleaner triggers happen, db2 will trigger dirty page
> threshold cleaner trigger, even the dirty pages in dirty queue have
> touched the page threshold???
> From my test, it looks like db2 does not use this way. But I am not
> sure. If db2 use is like this, how many number of dirty page steal
> cleaner trggers will cause the dirty page threshold cleaner trigger???
> Best Regards,
> Fan Ruo Xin


 
 
 

# of page steal cleaner will trigger page threshold?

Post by Fan Ruo Xi » Wed, 21 Aug 2002 11:32:26


Thanks!
Fan Ruo Xin

> No.

> Dirty steal triggers and dirty threshold triggers are completely
> different.  Threshold will only trigger at the change page threshold
> that you have setup in the db cfg.

> A dirty steal trigger will occur when a db2 agent victimizes
> a page which happens to be dirty.


> > Hello,
> > Is it db2 the same thing like informix in this way when a couple of
> > dirty page steal cleaner triggers happen, db2 will trigger dirty page
> > threshold cleaner trigger, even the dirty pages in dirty queue have
> > touched the page threshold???

> > From my test, it looks like db2 does not use this way. But I am not
> > sure. If db2 use is like this, how many number of dirty page steal
> > cleaner trggers will cause the dirty page threshold cleaner trigger???

> > Best Regards,
> > Fan Ruo Xin

 
 
 

1. Dirty page steal cleaner

I'm trying to gather some more information in regards to Dirty page
steal cleaner.

This is what we currently have from a snapshot:

Buffer pool data logical reads             = 277340825
Buffer pool data physical reads            = 2091285
Asynchronous pool data page reads          = 1169589
Buffer pool data writes                    = 209889
Asynchronous pool data page writes         = 203898
Buffer pool index logical reads            = 210378975
Buffer pool index physical reads           = 857713
Asynchronous pool index page reads         = 208049
Buffer pool index writes                   = 124775
Asynchronous pool index page writes        = 120361
Total buffer pool read time (ms)           = 6967517
Total buffer pool write time (ms)          = 123803
Total elapsed asynchronous read time       = 839026
Total elapsed asynchronous write time      = 48284
Asynchronous read requests                 = 159719
LSN Gap cleaner triggers                   = 769
Dirty page steal cleaner triggers          = 1553
Dirty page threshold cleaner triggers      = 0

Now my question is, from what point does one consider the Dirty page
steal cleaner triggers to be to high.
Also, I read that if it's too high, you increase the number of
num_iocleaners. But don't they just get triggered from the moment the
threshold is reached? And from what I can read here, that is never
reached. It also said that if "Buffer pool data writes" is almost
equal to "Asynchronous pool data page writes" to reduce the
num_iocleaners.
Another document stated that if this is too high then increase the
bufferpool. But our hits on the bufferpool is very good (over 99%).

As you can see, I'm a little bit confused about this and I can not
find any real good documentation on this.
So an help is appreaciated here on how to determine these things.

Thanks.

2. Going back to ado2.5 in win 2000

3. Stolen Pages Count

4. hard coded location?

5. Stolen page, In Beffer Manager Monitoring ...

6. Migrating from SQL Server 7 to Oracle8i

7. Stolen Page Count

8. DTS text file extract

9. High Stolen Page Count in PerfMon

10. Page cleaners / LRUS with single disk

11. Checkpoints too long, checking page cleaners !!

12. page cleaner #1 timed out

13. Page cleaners - have I understood correctly?