[Pdns-users] recursor and HUP
spork at bway.net
Wed Dec 26 05:26:48 UTC 2012
On Dec 25, 2012, at 11:38 PM, Charles Sprickman wrote:
> I'm not seeing anything in the documentation about what sending a HUP should do to the recursor, but here I'm seeing it kill the process.
> It gets a bit stranger than that though… We had a log rotation rule setup that was sending the recursor a HUP at midnight. This was unneeded since syslog is how pdns logs, so I'm not sure why we even had that there. That said, two identical boxes (same hardware, OS+version, same build of the same recursor pkg) were acting differently. The one that died on a HUP was where I was originally fiddling with controlsocket owner/location/mode, the one that did not die on a HUP was untouched, just running with defaults.
> Removing the socket after stopping the recursor and letting it set perms itself resulted in mode 700 and both now die on a HUP. However trying 750 and 755 is not "fixing" it. I'm suspicious of the socket owner/perms only because that's the only thing I know was originally different between the two boxes. Setting "trace=on" in the config file gives me no further information. In both cases the recursor just exits without logging anything.
> OS is FreeBSD 8.3, recursor is 3.3.
> What else can I look at to track this down?
Another datapoint… While looking at bugs in trac, I saw this one:
On a clean shutdown, I always see the recursor leaving a pidfile and control socket in /var/run after the process exits. Not sure if that's related at all, but it almost smells like something is a bit odd with signal handling.
> Charles Sprickman
> Bway.net - New York's Best Internet www.bway.net
> spork at bway.net - 212.655.9344
> Pdns-users mailing list
> Pdns-users at mailman.powerdns.com
More information about the Pdns-users