Rotating server logs?

[mage2k]

Is there any way to have a running server close and re-open it's log
file? Sending the server a SIGHUP signal seems to kill it outright.
On that last poing, should I open a bug report? A SIGHUP really
shouldn't be killing a server...

[Eliot Horowitz]

Are you using --logpath?
There isn't a way to do log rotation, but we could add.

You should add cases for sighup and log rotation.

[mage2k]

Thanks, will do.

[mage2k]

Ok, case opened for log rotation. I didn't, however, open a separate
case for SIGHUP signal handling since the default action for a process
is exit, which is what it does now, although I did suggest using it to
have the server re-open it's log in the log rotation case.

[Jon Hancock]

I'm not sure what SIGHUP should do. But if your going to have SIGHUP
help with log files per this requirement, then SIGHUP should probably
be doing things like re-reading config files as well.

Jon

[Eliot Horowitz]

What do you guys think about USR1 for rotating logs?
I'm a bti reluctant to use HUP

[timball]

i would say that USR1 would be the better signal to use. stays inline
w/ the nginx[0] way to do it. SIGHUP is used by apache[1] as a rather
large kludge, just remember that you'll need a global static volatile
sig_atomic_t somewhere to set the signal mask back[2] when coming back
from a function.

i can't believe you guys don't trap most usergenerated signals like
Richard W. Stevens says too!

but either way it would be *very* nice to be able to rotate a log. if
i get done making dinner early enough i might send in a patch tonight.

--timball

[0] http://www.ruby-forum.com/topic/134115
[1] http://httpd.apache.org/docs/2.2/stopping.html
[2] so says RWS in APUE 10.15

On Oct 15, 3:27 pm, Eliot Horowitz <eliothorow...@gmail.com> wrote:
> What do you guys think about USR1 for rotating logs?
> I'm a bti reluctant to use HUP
>