RE: Swap Space on DEC ALPHA with 7.30FC10 Engine
The virtual memory segment is sized so there is no segment allocations, we
only have our resident (Buffers) and a single virtual segment.
As for the size of swap space, this is a engine only server with 2GB of
ram, the entire engine runs within 875MB. This should allow close to 1GB
free, usually vmstat reports about 900MB free, although there are times that
it goes down to almost nothing. When inspecting the system, nothing
different seems to be running, the engine still only has it's 2 segments of
memory.
Is it possible that the engine could be swapping out part of virtual memory
when some other process runs on the UNIX server, but never releasing it
until the engine is bounced? What about setting the Forced residency flag
in the onconfig to 1, which will only have an effect of resident memory, not
virtual?
Thanks again,
John
-----Original Message-----
From: Dewhurst, Cy [mailto:cy.dewhurst@rbch-tr.swest.nhs.uk]
Sent: Thursday, February 08, 2001 9:30 AM
To: 'Porcello, John'
Subject: RE: Swap Space on DEC ALPHA with 7.30FC10 Engine
Rule of thumb for swap space configuration is normally 3 x physical
memory. 512k is very low.
I'd try increasing swap and monitoring shared memory segment allocation
with onstat -gr seg.
Try using onmode -F to free segments from time to time.
Immediate mode swap is more conservative than deferred but requires more
space. We use immediate here, so you could try linking
/sbin/swapdefault to your swap device, reboot and see if there is any
improvement.
I've heard of people having problems with deferred mode when swap runs
low but it may just be FUD, the manual pages for swapon certainly don't
indicate that it should cause any problems.
Overall, you may want to look at memory utilisation by each session.
onperf used to be a good tool for monitoring this, if you have dbcockpit
you could try that. Failing this, the onstat commands (onstat -g ses
and onstat -g ses SID), in the hope of pinning down what is responsible
for the leakage.
Cy
> -----Original Message-----
> From: Porcello, John [SMTP:jep-corp@kaman.com]
> Sent: Thursday, February 08, 2001 12:56 PM
> To: 'Dewhurst, Cy'
> Cc: Informix (E-mail)
> Subject: RE: Swap Space on DEC ALPHA with 7.30FC10 Engine
>
> The Primary Server has 2GB, the Secondary has 1GB, neither has
> /sbin/swapdefault, so we believe this means we use deferred mode.
> Also,
> they are both running Digital UNIX 4.0D, not 4.0G as originally
> stated.
> Both boxes have 4 processors, Primary box is a 4100 and the Secondary
> box is
> a 2100.
>
> Thanks,
> John
>
> -----Original Message-----
> From: Dewhurst, Cy [mailto:cy.dewhurst@rbch-tr.swest.nhs.uk]
> Sent: Thursday, February 08, 2001 6:42 AM
> To: 'Porcello, John'
> Subject: RE: Swap Space on DEC ALPHA with 7.30FC10 Engine
>
>
> How much physical memory do you have in each alpha? Is swap mode
> immediate or deferred (both boxes)?
>
> /sbin/swapdefault (exists?)
>
> Cy
>
> > -----Original Message-----
> > From: Porcello, John [SMTP:jep-corp@kaman.com]
> > Sent: Wednesday, February 07, 2001 9:13 PM
> > To: Informix (E-mail)
> > Subject: Swap Space on DEC ALPHA with 7.30FC10 Engine
> >
> > We are attempting to keep our engine running on a DEC Alpha with
> > Compaq Unix
> > 4.0G for a week without bouncing it. We have found that after about
> > three
> > days, our swap space starts getting utilized to the point that we
> need
> > to
> > bounce the engine. Once the engine is bounced, all swap space
> returns
> > and
> > we're good for another three days. This server is a Primary in a
> HDR
> > pair,
> > the Secondary server, doesn't seem to have the issue. This server
> > usually
> > has about 350 to 400 VERY active sessions running in an OLTP
> > environment.
> >
> > Any ideas would be greatly appreciated.
> >
> > Thanks,
> > John
> >
> > John E. Porcello, Jr., ICP, MCSE
> > Corporate DBA
> > Kaman Corporation
> > jep-corp@kaman.com
> >
> > My views and opinions may not reflect those of Kaman Corporation.
> >
> >
> > onstat -g glo> >
> > Informix Dynamic Server Version 7.30.FC10 -- On-Line (Prim) -- Up
> 2
> > days
> > 11:00:21 -- 856384 Kbytes
> >
> > MT global info:
> > sessions threads vps lngspins
> > 328 382 10 6
> >
> > sched calls thread switches yield 0 yield n yield
> > forever
> > total: 75571941 70081417 8018129 279769
> 29537656
> >
> > per sec: 5199 4120 1584 143 1948
> >
> >
> > Virtual processor summary:
> > class vps usercpu syscpu total
> > cpu 4 18252.63 6517.08 24769.72
> > aio 2 11.53 31.35 42.88
> > lio 1 0.87 2.62 3.48
> > pio 1 1.00 3.43 4.43
> > adm 1 0.35 0.75 1.10
> > msc 1 21.70 32.47 54.17
> > total 10 18288.08 6587.70 24875.78
> >
> > Individual virtual processors:
> > vp pid class usercpu syscpu total
> > 1 8817 cpu 4846.83 1771.32 6618.15
> > 2 8819 adm 0.35 0.75 1.10
> > 3 8809 cpu 4749.85 1674.17 6424.02
> > 4 8808 cpu 4535.37 1606.83 6142.20
> > 5 8823 cpu 4120.58 1464.77 5585.35
> > 6 8810 lio 0.87 2.62 3.48
> > 7 8806 pio 1.00 3.43 4.43
> > 8 8820 aio 8.23 16.85 25.08
> > 9 8825 msc 21.70 32.47 54.17
> > 10 8824 aio 3.30 14.50 17.80
> > tot 18288.08 6587.70 24875.78
> >
> > ps aux | grep oninit
> > USER PID %CPU %MEM VSZ RSS TTY S STARTED
> > TIME
> > COMMAND
> > root 8809 41.7 33.9 849M 693M ?? S Feb 05
> > 09:35:59
> > oninit
> > root 8817 30.9 34.6 849M 708M ?? R Feb 05
> > 09:00:25
> > oninit
> > root 8823 30.7 33.8 849M 692M ?? S Feb 05
> > 08:12:22
> > oninit
> > root 8808 28.5 33.8 849M 692M ?? R Feb 05
> > 09:08:03
> > oninit
> > root 8825 0.6 0.5 849M 10M ?? S Feb 05
> > 3:15.08
> > oninit
> > root 8820 0.2 0.2 849M 4.9M ?? S Feb 05
> > 1:31.91
> > oninit
> > root 8824 0.1 0.1 849M 2.6M ?? S Feb 05
> > 1:04.13
> > oninit
> > root 8810 0.0 0.0 849M 320K ?? S Feb 05
> > 0:13.06
> > oninit
> > root 8806 0.0 0.0 849M 336K ?? S Feb 05
> > 0:16.89
> > oninit
> > root 8819 0.0 0.4 849M 7.9M ?? S Feb 05
> > 0:07.14
> > oninit
> > root 24818 0.0 0.0 1.66M 184K ttyp3 S + 16:00:34
> > 0:00.01 grep
> > oninit> >
> >
> > onstat -g seg
> > Informix Dynamic Server Version 7.30.FC10 -- On-Line (Prim) -- Up> 2