Ugh, that's twice recently then. I need to revisit that pid handling code in MM. It
looks like it does open()/write()/close(), so not a lot of room for error, but who
knows...
--
Matt Domsch
Technology Strategist
Dell | Office of the CTO
-----Original Message-----
From: infrastructure-bounces(a)lists.fedoraproject.org
[mailto:infrastructure-bounces@lists.fedoraproject.org] On Behalf Of Ricky Zhou
Sent: Monday, November 15, 2010 3:33 AM
To: infrastructure(a)lists.fedoraproject.org
Subject: Re: Mirrormanager's crawler not running
On 2010-11-15 09:18:05 AM, Adrian Reber wrote:
According to a few hosts I checked in Mirrormanager's web
frontend the
crawler has not been running for at least three days. Last run was on
2010-11-12.
Hi, this was due to a stuck lockfile - I've restarted the crawler
now.
This happened earlier today with update-master-directory-list too, which caused people to
get hash mismatches with yum. What happened was that /var/run/mirrormanager/crawler.pid
and /var/run/mirrormanager/umdl.pid were empty, which prevented both of these processes
from running since sometime on the 13th.
I'm not sure how this condition occurs - maybe the process updating the pidfile
somehow got killed before it wrote the pid in? It is strange to see the exact same
problem with two different jobs though.
Thanks,
Ricky