Re: frequent 5.0_RC2 lockups cvs co / upd'ing to RF volume on SunFire v100

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Thu, Mar 19, 2009 at 03:18:44PM -0400, Rafal Boni wrote:
> I have a RF mirror over 2 IDE disks on my V100; I've moved a bunch of data
> off that mirror to an external NAS box, so I've got a lot of space I'd like
> to consolidate all various source trees to.
> 
> However, this has been very much less than succesfull; I've now hung the
> machine four or five times attempting to check out a fresh 'src' or 
> 'pkgsrc' from cvs.netbsd.org.  Each of these times, the machine was
> still responsive enough to get into ddb(4) but didn't answer pings
> and dropped all network connections.

Please make sure you have 
> 
> Some interesting tidbits that might be related:
>       * In at least 2 or 3 of the hangs, RF parity rebuild was likely
>         running.

Hum, remind me of what I did see on the old ftp.fr.netbsd.org:
any significant I/O with a SMP kernel while parity rebuild was running
would cause a hang. Using a UP kernel for parity rebuilds and switching
back to MP after a clean reboot would make the system work.
This was on i386 though, and with SCSI disks. I've never been able to
reproduce it on another system.
Anyway please make sure you have sys/dev/raidframe/rf_netbsdkintf.c 1.250.4.3
it may help ...

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--