Re: kill -2

On Wed, 2004-04-21 at 15:19, scott.marlowe wrote:
> Are you running the megaraid 2 driver or the older 1.18 series? Just
> wondering.

We started with the megaraid driver distributed with 2.6.3, which is
megaraid v2.00.3. With that driver we weren't able to insert more than 5
million row before corruption appeared. After upgrading to megaraid
v2.20.B2 and 2.6.5 things ran much better, but we are still able to get
the occational DB corruption, though we now have one table with 27
million rows which is working correctly, however, another smaller table
is still problematic. We were running XFS on LVM2 on linux RAID0 on
megaraid RAID1. I switched to JFS, and we are continuing to test.

I'm fairly confident that the problem is the bleeding edge 64bit compile
we are running, and not an underlying hardware problem, but I have to
look at everything.

> On Wed, 21 Apr 2004, Jord Tanner wrote:
>
> >
> > We are exploring the possibility of hardware problems, and have found
> > some issues with the LSILogic Megaraid driver running in 64bit (this is
> > a dual Opteron running Gentoo AMD64).
> >
> > Thanks for the feedback. I'm just covering all the bases.
> >
> > Jord Tanner
> >
> > On Wed, 2004-04-21 at 12:59, scott.marlowe wrote:
> > > On Wed, 21 Apr 2004, Jord Tanner wrote:
> > >
> > > >
> > > > Greetings,
> > > >
> > > > Does anyone know what the effect of sending an INT signal (kill -2) to
> > > > each of:
> > > > - pg backend
> > > > - psql running an insert or select into
> > > > - Perl script with DBI connection to pg, running insert or
> > > > select into
> > > > - vacuumdb
> > > >
> > > > I'm getting some database corruption, and I'm wondering if the use of
> > > > 'kill -2' could be the cause.
> > >
> > > No, kill -2 should not cause corruption.
> > >
> > > About once a month we get someone on the lists getting corruption. almost
> > > every time it's either a bad hard drive, memory, or system/cpu/mobo
> > > causing it. Have you tested your memory (memtest86) or drives (badblocks
> > > on linux) to make sure they aren't the cause of your problem? Just
> > > because a server is new doesn't mean it's free of all flaws.
> > >
> > >
> > > ---------------------------(end of broadcast)---------------------------
> > > TIP 7: don't forget to increase your free space map settings
> >
--
Jord Tanner <jord(at)indygecko(dot)com>