RE: AW: BUFFERS on Solaris-Box
- From: "Colin Dawson" <cjd_1955@xxxxxxxxxxx>
- Date: Fri, 24 Feb 2006 11:24:42 +0000
I run IDS 7.31.FD7 on a Solaris 9 server,
You are running a very old patch level of Solaris - install the latest 9_Recommended patch cluster. It won't help you with this issue, but good from a server management perspective :)
I have
RESIDENT -1
BUFFERS 1900000 (4Gb Solaris SHM Segment limitation)
SHMVIRTSIZE 4096000
SHMADD 4014080
and I normally get 2 additional segments added within the 2 hours of starting the DB, I also normally have approx 1400 sessions connected
Search CDI Archives for a discussion on this topic, I think it was probably 18 months ago
Regards
Colin
There are 10 types of people in the world, those that understand binary and those that don't
From: "Habichtsberg, Reinhard" <RHabichtsberg@xxxxxxxxxxxxxxxxxx>
To: "Informix-List (E-Mail)" <informix-list@xxxxxxxx>, "'Chafik, Basim'"<Basim.Chafik@xxxxxxxxxx>
Subject: AW: BUFFERS on Solaris-Box
Date: Thu, 23 Feb 2006 17:10:54 +0100
Hi Basim,
thanks for your fast answer.
I agree with you in gradually tuning the parameters though it's a bit
difficult because both systems are running 7x24.
onstat -a produces files which have - compressed by zip - a size of 10 MB
and 5 MB. Are you sure I should send them?
Another question arised: We configured 350.000 buffers and buffsize is
2048. So I expected the size of resident portion about 700 MB.
onstat -g seg shows the following:
Segment Summary:
id key addr size ovhd class blkused
blkfree
1 1381451777 20000000 1279262720 457728 R 312281
39
2 1381451778 6c400000 503316480 16016 V 65108
57772
3 1381451779 8a400000 3145728 752 M 530
238
4 1381451780 8a700000 3145728 752 M 528
240
8 1381451781 8aa00000 100663296 3728 V 691
23885
9 1381451782 90a00000 100663296 3728 V 3209
21367
10 1381451783 96a00000 100663296 3728 V 713
23863
The size of the resident portion seems to be 1220 MB. I would be glad if you
could explain this behaviour.
Thanks,
Reinhard
> -----Ursprüngliche Nachricht-----
> Von: Chafik, Basim [mailto:Basim.Chafik@xxxxxxxxxx]
> Gesendet am: Donnerstag, 23. Februar 2006 15:20
> An: 'Habichtsberg, Reinhard'
> Betreff: RE: BUFFERS on Solaris-Box
>
> Hi Reinhard;
>
> More is not always better. Sometime you give informix
> resources that will be
> waisted. Those resources can be used elsewhere like users
> process. In your
> case you certinly need more buffers but I would not go from 350000 to
> 2500000 in one step. What I would do usually is to go
> gradually, and then
> monitor (that is why it is called tuning). I will double the
> buffers , and
> then I will monitor for week to see how are my stats (onstat
> -p) are lloking
> now. Then I f it still need more I will give more and monitor
> again until I
> reach the magic number.
>
> For LRUs and cleaner and the other question, more info will be helpful
> (nobody can just give you a magic fomula for you system). If
> you send the
> 'onstat -a' that would be helpful.
>
> Thanks
>
> Basim Chafik
> Senior Systems Analyst
> IBM Certified Advanced Technical Expert (CATE)
> 1.800.688.4895
> basim.chafik@xxxxxxx
> plexus (Division of BancTec)
>
>
>
> -----Original Message-----
> From: Habichtsberg, Reinhard [mailto:RHabichtsberg@xxxxxxxxxxxxxxxxxx]
> Sent: Thursday, February 23, 2006 9:03 AM
> To: Informix-List (E-Mail)
> Subject: BUFFERS on Solaris-Box
>
>
> Hi,
>
> new job, new environment.
>
> I found two Solaris Database Server with IDS 9.40.FC4W2 here.
> The one has 32
> GB RAM the other 12 GB RAM.
> Server db1 (32 GB) Uname -a: SunOS db1 5.9 Generic_117171-12
> sun4u sparc
> SUNW,Netra-T12
> Server db2 (12 GB) Uname -a: SunOS db2 5.8 Generic_108528-29
> sun4us sparc
> FJSV,GPUSC-M
>
> Both systems are under heavy load. Monitoring the waitstates
> shows that AIO
> und BUFFER waits are higher than running. onstat -p show 92 %
> cached reads
> and only 50 % cache writes on the bigger system and 96 % to
> 83 % on the
> smaller one.
>
> The onconfig param BUFFER is set to 350.000 (700 MB) in both
> systems. In my
> opinion the systems would work faster with a much higher
> amount of buffers
> (for example 2.500.000 =5 GB).
>
> The bigger system shows up to 1500 user sessions the smaller
> up 400. In both
> system the SHMVIRTSIZE is 491520 and SHMADD is 98304. The
> bigger system
> really reserves additional virtual memory segments when
> running. The amount
> of DATA is about 1 TB for each server.
>
> What value do you suggest for the BUFFERS (and CLEANERS,
> LRUS, LRU_MAX_DIRTY
> and LRU_MAX_DIRTY)? Are there any restrictions with _Solaris_?
>
> Any suggestions are welcome.
>
> TIA
> Reinhard.
>
> _______________________________________________
> Informix-list mailing list
> Informix-list@xxxxxxxx
> http://www.iiug.org/mailman/listinfo/informix-list
>
_______________________________________________
Informix-list mailing list
Informix-list@xxxxxxxx
http://www.iiug.org/mailman/listinfo/informix-list
.
- Prev by Date: Re: AW: BUFFERS on Solaris-Box
- Next by Date: RE: New Repostory Contribution - PHP PDO INFORMIX
- Previous by thread: Re: AW: BUFFERS on Solaris-Box
- Next by thread: Unloading Module <SPLNULL> LOading Module <SPLNULL>
- Index(es):
Relevant Pages
|
