Eric Sandeen wrote:
>
> Jim Eshleman wrote:
>
> > In my case the "Bad write on page" always preceeds "Warning: buffer
> > ... with weird blockno" and then a pair of "kernel BUG at ll_rw_blk.c"
> > errors. Attached is the syslog output from the most recent, 2.4.9 +
> > linux-2.4.9-xfs-2001-08-26.patch. Any ideas if this is XFS, 2.4 VM bug,
> > something else?
>
> Hm.... ok I have a 2.4.9 system up and running now, what do you do to
> hit this error? Is it easy to reproduce?
Okay, get about 5,000 users or so to hit it with POP/IMAP connections
and use it as their incoming/outgoing SMTP server :-) No, sorry, it's
not easy to reproduce. I've not been able to pinpoint anything that
triggers the error. It has happened once early in the AM during
automated backup, and twice in the evening (5-7PM) when the load is
moderate. It has not happened during working hours yet, when load is
highest.
From my previous e-mail in this thread:
> I also am seeing the "Bad write on page" errors. Attached is syslog
> info from two incidents. The first was after three weeks uptime, the
> second was after one week uptime. The system is a heavily-loaded mail
> server, IBM Netfinity x370 8-way PIII Xeon, 8.5GB RAM (HIGHMEM enabled)
> running 2.4.9 + patch-2.4.9-xfs-2001-08-17 compiled with egcs-2.91.66.
> The XFS filesystems are on top of LVM, on top of hardware RAID1E. After
> the errors the system becomes unresponsive. I suspected a 2.4 VM bug
> but searching LKML archives turns up no hits, so maybe it is XFS
> related. I also saw the error after six hours running 2.4.9 +
> linux-2.4.9-xfs-2001-08-26.patch.
Jim
|