innd stops responding
Heath Kehoe
hakehoe at avalon.net
Mon Jul 19 14:39:59 UTC 2004
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Jul 19, 2004, at 8:50, Alex Kiernan wrote:
> On Mon, 19 Jul 2004 06:42:48 -0700, Gary Thandi <gary.thandi at 360.net>
> wrote:
>>
>> yes. i've had to resort to stopping and starting all news processes
>> every
>> 60 minutes.
>>
>
> Try getting a snapshot - I know Heath's made some fixes to the ovdb
> code, but I've not tried those.
Either that, or you can get this patch:
http://www.avalon.net/~hakehoe/inn/ovdb_2.4.1_patch.txt
> Certainly the behaviour you're seeing is typical for ovdb on Solaris;
> I'm not sure if anyone's addressed the way nnrpd does termination yet,
> but last time I looked it seemed like if you took a signal at the
> wrong moment, nnrpd would try and make calls into ovdb (and hence
> BerkeleyDB) in a re-entrant fashion, which isn't supported (not
> unreasonably) by BerkeleyDB.
Whoa!!! I never knew about this. OK, looks like the sigpipe handler
calls ExitWithStats. Yuck!! But how often does nnrpd exit in this way?
There's an easy workaround: turn on readserver in ovdb.conf, and
restart
news. With readserver on, the nnrpd doesn't access BerkeleyDB directly,
thus avoiding the signal handler problem.
- - h
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)
iD8DBQFA+91O4uXPAG0A1J4RAkw4AJ4/80HO4vQjnLD3KgFUUMMy1x2rSACeM3xD
IJautqYW8B/CGhBvy37SZWw=
=xA3i
-----END PGP SIGNATURE-----
More information about the inn-workers
mailing list