Hello everyone.
I'm experiencing Informix behaviour that feels weird to me.
Sometimes, let's say once a week on average, read cache rate drops suddenly.
Most of the time read cache rate is very high, around 99.7 percent. And all
of the sudden it drops to somewhere around 50%.
Then it spends few hours reaching 95% and a day or two to get over 99%
level. As you've already guessed, after few more days the story repeats
itself.
It never happens with write cache rate, though.
I never experience buffer overflow, that is "onstat -p" reports ovbuff to be
zero, so I guess I have enough buffers configured.
Config is: IDS 9.40UC5W2 on Solaris 9, Sun Fire V480 with 2 processors and 4
GB RAM.
Thanks everyone.
Christian Knappke - 30 Jun 2005 15:22 GMT
From the keyboard of "Davorin Kremenjas"
<davorin.kremenjas@srce.hr>:
> Sometimes, let's say once a week on average, read cache rate
> drops suddenly. Most of the time read cache rate is very high,
[quoted text clipped - 7 lines]
>
> Config is: IDS 9.40UC5W2 on Solaris 9, Sun Fire V480 with 2
. ^^^^^^^^^
This is 32 bits -----|
> processors and 4 GB RAM.
In 32 bit servers most counters are also 32 bit and roll over when
they reach 2^32-1 (4294967295) or 2^31-1 (2147483647). If that
happens, then the derived value for read cache hit rate is
meaningless. It is even meaningless for the rest of the database run
time, because it depends on two values (dskreads and bufreads) that
grow at different rates. To get the value back to the level of truth
you need to use "onstat -z".
Best regards
Christian

Signature
#include <std_disclaimer.h>
/* The opinions stated above are my own and not
necessarily those of my employer. */
Davorin Kremenjas - 30 Jun 2005 15:48 GMT
> In 32 bit servers most counters are also 32 bit and roll over when
> they reach 2^32-1 (4294967295) or 2^31-1 (2147483647). If that
I suppose this is the case. At this moment buffreads is around 2^30, it will
reach 2^31 or 2^32 sometimes soon since it grows quite fast over time.
Thanks Christian, mistery solved.
Bill Dare - 30 Jun 2005 15:50 GMT
Are you ever resetting statistics with onstat -z? Sounds like you are
intermittently hitting a point where some value wraps back to zero
causing the cache hit rate to be inaccurate.
Bill
> -----Original Message-----
> From: owner-informix-list@iiug.org [SMTP:owner-informix-list@iiug.org]
[quoted text clipped - 25 lines]
>
> Thanks everyone.
sending to informix-list
Zev Berezin - 30 Jun 2005 17:19 GMT
It is possible that you run a DSS type report every once
in a while and that would bring the read cache rate down.
HTH,
Zev Berezin
Database Administrator
B&H Photo - www.bhphoto.com
From: "Davorin Kremenjas" <davorin.kremenjas@srce.hr>
To: informix-list@iiug.org
Subject: Sudden drop in read cache rate - why?
Date sent: Thu, 30 Jun 2005 13:36:06 +0200
Organization: CARNet, CROATIA
Send reply to: "Davorin Kremenjas" <davorin.kremenjas@srce.hr>
> Hello everyone.
> I'm experiencing Informix behaviour that feels weird to me.
[quoted text clipped - 12 lines]
>
> Thanks everyone.
sending to informix-list