Re: benchmark woes and XFS options - Mailing list pgsql-performance

From mark
Subject Re: benchmark woes and XFS options
Date
Msg-id 013001cc564a$c44d7140$4ce853c0$@com
Whole thread Raw
In response to Re: benchmark woes and XFS options  (Greg Smith <greg@2ndquadrant.com>)
List pgsql-performance

> -----Original Message-----
> From: Greg Smith [mailto:greg@2ndquadrant.com]
> Sent: Monday, August 08, 2011 9:42 PM
> To: mark
> Cc: pgsql-performance@postgresql.org
> Subject: Re: [PERFORM] benchmark woes and XFS options
>
> I think your notion that you have an HP CCISS driver in this older
> kernel that just doesn't drive your card very fast is worth exploring.
> What I sometimes do in the situation you're in is boot a Linux
> distribution that comes with a decent live CD, such as Debian or
> Ubuntu.  Just mount the suspect drive, punch up read-ahead, and re-test
> performance.  That should work well enough to do a simple dd test, and
> probably well enough to compile and run bonnie++ too.  If that gives
> good performance numbers, it should narrow the list of possible causes
> considerably.  You really need to separate out "bad driver" from the
> other possibilities here given what you've described, and that's a low
> impact way to do it.
>
> --
> Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
> PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us

Thanks Greg.

I will try and give HPSA a whirl and report back. Both with single disk and
the whole raid set. I am out of the office this week so I might have some
delay before I can do some more detective work.

I don't think I have any gear that won't require either the CCISS or HPSA
driver and in SFF drives. But will try and look around.



-Mark




pgsql-performance by date:

Previous
From: Greg Smith
Date:
Subject: Re: Suspected Postgres Datacorruption
Next
From: Sumeet Jauhar
Date:
Subject: Re: Suspected Postgres Datacorruption