[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Sks-devel] SKS 1 minute timeouts
From: |
Jonathon Weiss |
Subject: |
Re: [Sks-devel] SKS 1 minute timeouts |
Date: |
Mon, 13 Jul 2009 19:32:13 -0400 |
> > Having just upgraded pgp.mit.edu to SKS, I've been paying closer than
> > usual attention to it and, at least my nagios installation, has seen
> > some problems. Those problems seem to correlate with 1 minute empty
> > spaces in db.log. Sometimes they include log messages about timeouts,
> > eg:
> >=20
> > >> 2009-07-07 14:17:31 /pks/lookup: Index request: (0x3f3e6426)
>
> Esp. for users still on dialup, it can take a while to transfer larger
> keys like that one:
>
> %gpg --recv 0x3f3e6426
> %gpg --export -a 0x3f3e6426 | wc -c
> 144639
>
> > >> user-agent:Mozilla/4.0 (compatible; MSIE 7.0b; Windows NT 6.0)]): No k=
> eys found
> >=20
> > or probably some other things
> >=20
> > And sometimes there's no error, just two consecutive log entries
> > exactly one minute apart.
> >=20
> > I know htat the machine will spend a certain amount of time in I/O
> > wait when it's grovelling the database, but that tends to be seconds
> > at a time, and the fact that these all seem to be exactly a minute
> > sounds like some sort of internal timeout to me. Has anyone else seen
> > this or have any idea what could be causing it?
>
> I think that is set here:
>
> settings.ml:173:let command_timeout =3D ref 60
>
> Note that SKS doesn't multiplex its I/O, so the "sks db" daemon
> should be tied-up/unresponsive to other clients during these time[out]s,
> IIRC. Therefore increasing the timeout probably isn't desirable.
Yeah, I was actually tempted to *reduce* it, though I'm not sure how
badly that would hurt dialup users.
Jonathon
Jonathon Weiss <address@hidden>
MIT/IS&T/OIS Server Operations