How to reduce bufwaits in IDS?
A DBA running IDS 7.31 on Digital Unix with four PeopleSoft Financials databases reported a very high bufwaits count (~7.9M, about 30% of buffer writes) in onstat -p and asked how to reduce it. Respondents (David Williams, Art Kagel) recommended raising LRUS and CLEANERS to the maximum of 128, monitoring the cleaners with onstat -u to see how many stay idle and tuning down accordingly, increasing BUFFERS if memory allows, and looking at LRU_MAX_DIRTY/LRUPRIORITY guidance in the FAQ. Others asked for onstat -c and onstat -g spi output for further diagnosis. The poster never followed up, so no confirmed outcome is recorded.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Hello,
I'm seeing a problem with my IDS instance (7.31 for Digital Unix 4.0)
which contains 4 PeopleSoft Financials databases (about 400-600 MB each)
There are a large number of bufwaits when I run the onstat command.
See results below.
Can anyone suggest ways to reduce this high figure?
Informix Dynamic Server Version 7.31.FC4A -- On-Line -- Up 30 days
05:42:32 -- 458752 Kbytes
Profile
dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
27247923 17395712 2555873294 98.93 915986 2423007 4185287 78.11
isamtot open start read write rewrite delete
2329311336 50483831 118027700 1601044582 962372 359135 177949
commit rollbk
220032 132
gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
0 0 0 0 0 0 0
ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
0 0 0 215885.50 44764.13 8722 17444
bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
7892336 82 2565639368 0 0 367 54473 2646791
ixda-RA idx-RA da-RA RA-pgsused lchwaits
852324 5813854 13380955 19961138 91
Thanks for any help.
Regards,
Samuel Lee
Trinidad.
Sent via Deja.com http://www.deja.com/
Before you buy.
Sorry the message got scrambled.
This is much better:
Hello,
I'm seeing a problem with my IDS instance (7.31 for Digital Unix 4.0)
which contains 4 PeopleSoft Financials databases (about 400-600 MB
each)
There are a large number of bufwaits when I run the onstat command. See
results below.
Can anyone suggest ways to reduce this high figure?
Informix Dynamic Server Version 7.31.FC4A -- On-Line -- Up 30 days
05:42:32 -- 458752 Kbytes
Profile
dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
27247923 17395712 2555873294 98.93 915986 2423007 4185287 78.11
isamtot open start read write rewrite delete
2329311336 50483831 118027700 1601044582 962372 359135 177949
commit rollbk
220032 132
gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
0 0 0 0 0 0 0
ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
0 0 0 215885.50 44764.13 8722 17444
bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
7892336 82 2565639368 0 0 367 54473 2646791
ixda-RA idx-RA da-RA RA-pgsused lchwaits
852324 5813854 13380955 19961138 91
Thanks for any help.
Regards,
Samuel Lee
Trinidad.
Sent via Deja.com http://www.deja.com/
Before you buy.
Sorry the message got scrambled.
This is much better:
Hello,
I'm seeing a problem with my IDS instance (7.31 for Digital Unix 4.0)
which contains 4 PeopleSoft Financials databases (about 400-600 MB
each)
There are a large number of bufwaits when I run the onstat command. See
results below.
Can anyone suggest ways to reduce this high figure?
Informix Dynamic Server Version 7.31.FC4A -- On-Line -- Up 30 days
05:42:32 -- 458752 Kbytes
Profile
dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
27247923 17395712 2555873294 98.93 915986 2423007 4185287 78.11
isamtot open start read write rewrite delete
2329311336 50483831 118027700 1601044582 962372 359135 177949
commit rollbk
220032 132
gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
0 0 0 0 0 0 0
ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
0 0 0 215885.50 44764.13 8722 17444
bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
7892336 82 2565639368 0 0 367 54473 2646791
ixda-RA idx-RA da-RA RA-pgsused lchwaits
852324 5813854 13380955 19961138 91
Thanks for any help.
Regards,
Samuel Lee
Trinidad.
Sent via Deja.com http://www.deja.com/
Before you buy.
Increase LRUS to 128
increase CLEANERS to 128
monitor with onstat -u (cleaners are near the start, all with similar
flags.
See how many cleaners stay idle after a day (all 0's for operations).
Tune accordingly.
If this fails increase BUFFERS. Or look at the new FAQ
at www.smooth1.demon.co.uk for the new section on LRUPRIORITY
David.
sleesingchan@my-deja.com wrote in message <8o41e5$gfl$1@nnrp1.deja.com>...
>Sorry the message got scrambled.
>This is much better:
>
>Hello,
>
>I'm seeing a problem with my IDS instance (7.31 for Digital Unix 4.0)
>which contains 4 PeopleSoft Financials databases (about 400-600 MB
>each)
>
>There are a large number of bufwaits when I run the onstat command. See
>results below.
>
>Can anyone suggest ways to reduce this high figure?
>
>Informix Dynamic Server Version 7.31.FC4A -- On-Line -- Up 30 days
>05:42:32 -- 458752 Kbytes
>
>Profile
>dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
>27247923 17395712 2555873294 98.93 915986 2423007 4185287 78.11
>
>isamtot open start read write rewrite delete
>2329311336 50483831 118027700 1601044582 962372 359135 177949
>commit rollbk
>220032 132
>
>gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
>0 0 0 0 0 0 0
>
>ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
>0 0 0 215885.50 44764.13 8722 17444
>
>bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
>7892336 82 2565639368 0 0 367 54473 2646791
>
>ixda-RA idx-RA da-RA RA-pgsused lchwaits
>852324 5813854 13380955 19961138 91
>
>Thanks for any help.
>Regards,
>Samuel Lee
>Trinidad.
>
>
>Sent via Deja.com http://www.deja.com/
>Before you buy.
Yup a Bufwaits Ratio of 30% is high. Increase LRUS and CLEANERS to the max
value of 128, increase BUFFERS if you have the memory also.
Art S. Kagel
sleesingchan@my-deja.com wrote:
>
> Sorry the message got scrambled.
> This is much better:
>
> Hello,
>
> I'm seeing a problem with my IDS instance (7.31 for Digital Unix 4.0)
> which contains 4 PeopleSoft Financials databases (about 400-600 MB
> each)
>
> There are a large number of bufwaits when I run the onstat command. See
> results below.
>
> Can anyone suggest ways to reduce this high figure?
>
> Informix Dynamic Server Version 7.31.FC4A -- On-Line -- Up 30 days
> 05:42:32 -- 458752 Kbytes
>
> Profile
> dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
> 27247923 17395712 2555873294 98.93 915986 2423007 4185287 78.11
>
> isamtot open start read write rewrite delete
> 2329311336 50483831 118027700 1601044582 962372 359135 177949
> commit rollbk
> 220032 132
>
> gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
> 0 0 0 0 0 0 0
>
> ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
> 0 0 0 215885.50 44764.13 8722 17444
>
> bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
> 7892336 82 2565639368 0 0 367 54473 2646791
>
> ixda-RA idx-RA da-RA RA-pgsused lchwaits
> 852324 5813854 13380955 19961138 91
>
> Thanks for any help.
> Regards,
> Samuel Lee
> Trinidad.
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
It might be helpful if we has the output from onstat -c as well
Will
sleesingchan@my-deja.com wrote:
> Sorry the message got scrambled.
> This is much better:
>
> Hello,
>
> I'm seeing a problem with my IDS instance (7.31 for Digital Unix 4.0)
> which contains 4 PeopleSoft Financials databases (about 400-600 MB
> each)
>
> There are a large number of bufwaits when I run the onstat command. See
> results below.
>
> Can anyone suggest ways to reduce this high figure?
>
> Informix Dynamic Server Version 7.31.FC4A -- On-Line -- Up 30 days
> 05:42:32 -- 458752 Kbytes
>
> Profile
> dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
> 27247923 17395712 2555873294 98.93 915986 2423007 4185287 78.11
>
> isamtot open start read write rewrite delete
> 2329311336 50483831 118027700 1601044582 962372 359135 177949
> commit rollbk
> 220032 132
>
> gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
> 0 0 0 0 0 0 0
>
> ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
> 0 0 0 215885.50 44764.13 8722 17444
>
> bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
> 7892336 82 2565639368 0 0 367 54473 2646791
>
> ixda-RA idx-RA da-RA RA-pgsused lchwaits
> 852324 5813854 13380955 19961138 91
>
> Thanks for any help.
> Regards,
> Samuel Lee
> Trinidad.
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
Of course it might be helpful if I could type too.
Anyway, onstat -c would let us look at how you had the system configured.
onstat -g spi | sort -rn +1 | head -20might be interesting as well.
William Rice wrote:
> It might be helpful if we has the output from onstat -c as well
>
> Will
>
> sleesingchan@my-deja.com wrote:
>
> > Sorry the message got scrambled.
> > This is much better:
> >
> > Hello,
> >
> > I'm seeing a problem with my IDS instance (7.31 for Digital Unix 4.0)
> > which contains 4 PeopleSoft Financials databases (about 400-600 MB
> > each)
> >
> > There are a large number of bufwaits when I run the onstat command. See
> > results below.
> >
> > Can anyone suggest ways to reduce this high figure?
> >
> > Informix Dynamic Server Version 7.31.FC4A -- On-Line -- Up 30 days
> > 05:42:32 -- 458752 Kbytes
> >
> > Profile
> > dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
> > 27247923 17395712 2555873294 98.93 915986 2423007 4185287 78.11
> >
> > isamtot open start read write rewrite delete
> > 2329311336 50483831 118027700 1601044582 962372 359135 177949
> > commit rollbk
> > 220032 132
> >
> > gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
> > 0 0 0 0 0 0 0
> >
> > ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
> > 0 0 0 215885.50 44764.13 8722 17444
> >
> > bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
> > 7892336 82 2565639368 0 0 367 54473 2646791
> >
> > ixda-RA idx-RA da-RA RA-pgsused lchwaits
> > 852324 5813854 13380955 19961138 91
> >
> > Thanks for any help.
> > Regards,
> > Samuel Lee
> > Trinidad.
> >
> > Sent via Deja.com http://www.deja.com/
> > Before you buy.