Date: Fri, 5 May 2000 09:12:09 +0930 From: Greg Lehey <grog@lemis.com> To: Thomas Faehnle <tf@wurbl.wn.bawue.de> Cc: questions@FreeBSD.ORG Subject: Re: 4.0-RELEASE: vinum, fs, ATA trouble Message-ID: <20000505091209.D32650@freebie.lemis.com> In-Reply-To: <E12nLgj-0002Ia-00@wurbl.bk.int> References: <E12nLgj-0002Ia-00@wurbl.bk.int>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday, 4 May 2000 at 15:27:26 +0200, Thomas Faehnle wrote:
> Hi all,
>
> I'm currently evaluating the 4.0-RELEASE for use as a fileserver in a
> production environment.
>
> The desired configuration includes vinum in RAID5 mode. Soft updates
> would be Nice To Have, but this isn't a must if they're still
> problematic in combination with vinum[1].
>
> My own tests were quite discouraging. I was able to provoke kernel
> error messages ("dscheck(#ad/0x20014): b_bcount 8 is not on a sector
> boundary (ssize 512)"), corrupted files, and kernel panics
> ("ffs_valloc: dup alloc") within a few minutes by simply copying
> the /usr/src tree to a vinum volume that had one of its plexes revived
> at that time[2]. Soft updates were enabled on the vinum filesystem.
Interesting. I've only been able to reproduce the panics.
> But then, copying /usr/src to a filesystem with non-default block and
> fragment size (32k/4k) yields the cp process hanging in "D+" state
> after some 10s of megabytes. No vinum, no soft updates involved.
More interesting than the D+ state is the WCHAN value returned by
ps l.
> [1] BUGS section of vinum(4): "... Some problems have been reported
> with vinum in combination with soft updates."
Yes, this is one of them. The problems were particularly associated
with IDE drives. I'm currently testing a fix, and it looks good. It
should be available in a day or two.
> [2] The vinum config was:
>
> drive d0 device /dev/ad2s1e
> drive d1 device /dev/ad2s2e
> drive d2 device /dev/ad2s3e
> drive d3 device /dev/ad2s4e
It makes no sense at all to have more than one drive per spindle. You
just waste space, and configuration updates take longer.
> volume perverse
> plex org raid5 256k
> subdisk len 100m drive d0
> subdisk len 100m drive d1
> subdisk len 100m drive d2
> subdisk len 100m drive d3
> plex org striped 256k
> subdisk len 150m drive d0
> subdisk len 150m drive d1
> plex org concat
> subdisk len 180m drive d2
> subdisk len 120m drive d3
The rest of the config looks OK. But you only need one drive. Here's
the config I'm testing on at the moment:
# Vinum configuration of panic.lemis.com, saved at Fri May 5 09:10:20 2000
drive ata device /dev/ad0s2h
volume sos
plex name sos.p0 org raid5 512s vol sos
sd name sos.p0.s0 drive ata plex sos.p0 len 1638400s driveoffset 2313s plexoffset 0s
sd name sos.p0.s1 drive ata plex sos.p0 len 1638400s driveoffset 1640713s plexoffset 512s
sd name sos.p0.s2 drive ata plex sos.p0 len 1638400s driveoffset 3279113s plexoffset 1024s
sd name sos.p0.s3 drive ata plex sos.p0 len 1638400s driveoffset 4917513s plexoffset 1536s
sd name sos.p0.s4 drive ata plex sos.p0 len 1638400s driveoffset 6555913s plexoffset 2048s
sd name sos.p0.s5 drive ata plex sos.p0 len 1638400s driveoffset 8194313s plexoffset 2560s
sd name sos.p0.s6 drive ata plex sos.p0 len 1638400s driveoffset 9832713s plexoffset 3072s
Greg
--
When replying to this message, please copy the original recipients.
For more information, see http://www.lemis.com/questions.html
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000505091209.D32650>