Randy Bush wrote:
>> I smell a corrupted DBM file. Can you run exim_tidydb by hand (it
>> won't hurt) at an appropriate moment and wait to see which db file it
>> opens when the memory usage goes bananas?
>
> work0.psg.com:/var/spool/exim/db# exim_tidydb /var/spool/exim retry
> Tidying Exim hints database /var/spool/exim/db/retry
> failed to malloc 8200 bytes of memory: called from line 167 of store.c
>
> work0.psg.com:/var/spool/exim/db# ls -l
> total 128
> -rw-r----- 1 mailnull mail 65536 Apr 27 21:46 retry
> -rw-r----- 1 mailnull mail 0 May 25 2003 retry.lockfile
> -rw-r----- 1 mailnull mail 65536 Apr 27 21:55 wait-remote_smtp
> -rw-r----- 1 mailnull mail 0 May 25 2003 wait-remote_smtp.lockfile
>
> work0.psg.com:/var/spool/exim/db# rm *
>
> wait a while
>
> work0.psg.com:/var/spool/exim/db# ls -l
> total 35
> -rw-r----- 1 mailnull mail 16384 Apr 27 22:20 retry
> -rw-r----- 1 mailnull mail 0 Apr 27 22:10 retry.lockfile
> -rw-r----- 1 mailnull mail 16384 Apr 27 22:25 wait-remote_smtp
> -rw-r----- 1 mailnull mail 0 Apr 27 22:10 wait-remote_smtp.lockfile
>
> work0.psg.com:/var/spool/exim/db# exim_tidydb /var/spool/exim retry
> Tidying Exim hints database /var/spool/exim/db/retry
>
>
> top shows
>
> PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND
> 38730 root 1 62 0 7772M 3432M pfault 1 0:25 16.26% exim_tidydb
>
>
> system locks up and requires remote power bar
>
> this is happening on more than one system
>
> randy
>
- the OS is the same production stable version on all of them?
or NOT [1]
..and
- exim was compiled on the same version and the very same box it is
running on?
or NOT [2]
'The World Wonders'
Well actually not. See notes..
Bill Hacker
[1] For FreeBSD, that means not later than 7.1 RELEASE.
FreeBSD 8-CURRENT AMD64 does NOT qualify.
CURRENT is specified to be unsuited for production use, and RELENG can
be trusted on that score.
Even 7.2-RC(x) is still a bit of a dice-roll. BTDTGTTS,WBH [4]. Closet
full of 'em.
Not to put too fine a point on it, but even ~STABLE branch has had the
odd careless MFC the past several years.
Further, AFAICS, this specific problem has not been reported, on *any*
Linux (yet), nor even other *BSD releases/variant.
IOW - does not appear to be an Exim issue.
[2] AND NOT installed from a binary or binary package built on someone
else's box with God-alone knows what glitches in libs?. Or even not-64-bit?
[3] JFWIW, OpenBSD by comparison has been rock-solid the past 2+ years
even as binary-applied daily snapshots of their 'CURRENT' branch. Tight
code-auditing produces more than just good security.
OTOH, while getting faster at each release, OpenBSD will generally
'ubench' or 'bytebench' the same hardware about half as fast as FreeBSD.
[4] Been There, Done That, Got The Tee Shirt, With Bullet Holes!
One makes choices.....
