You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by sameer <sa...@c2.net> on 1997/11/07 00:33:56 UTC
Re: general/1376: Server stops servicing requests when log disk is full (fwd)
> Erm.... exactly what are we supposed to do here?
>
> rm -Rf / ?
stop logging?
--
Sameer Parekh Voice: 510-986-8770
President FAX: 510-986-8777
C2Net
http://www.c2.net/ sameer@c2.net
Re: general/1376: Server stops servicing requests when log disk is full (fwd)
Posted by Dean Gaudet <dg...@arctic.org>.
Using a loopback device mount under linux (treats a file as a device) I
created a 256k filesystem, and started apache with error log and
access_log on that filesystem. Then I used dd to fill the disk. Then I
hit it with a lot of hits and traced the children, I kept going until the
children were getting ENOSPC for both error_log and access_log writes.
Nothing ill happened.
The only issue is that Apache makes no attempt to report these errors...
it could report ENOSPC once per child per access_log. But it can't easily
report error_log ENOSPC. It could report them if syslog is active. At
any rate I'm not likely to make a patch for this. But if someone does
make sure you rate limit it to one report per log per child.
Dean
On Thu, 6 Nov 1997, Dean Gaudet wrote:
> None of the logging code actually checks the return value from the various
> write functions. But that shouldn't stop it from serving hits. I suspect
> something else is borken. Like, for example, they have a CGI which blocks
> badly on full disk and that's chewing up all their children.
>
> Or perhaps the lockfile can't be locked on a full disk? That'd be cute.
>
> Dean
>
> On Thu, 6 Nov 1997, sameer wrote:
>
> > > Erm.... exactly what are we supposed to do here?
> > >
> > > rm -Rf / ?
> >
> > stop logging?
> >
> > --
> > Sameer Parekh Voice: 510-986-8770
> > President FAX: 510-986-8777
> > C2Net
> > http://www.c2.net/ sameer@c2.net
> >
>
>
Re: general/1376: Server stops servicing requests when log disk is full (fwd)
Posted by Dean Gaudet <dg...@arctic.org>.
None of the logging code actually checks the return value from the various
write functions. But that shouldn't stop it from serving hits. I suspect
something else is borken. Like, for example, they have a CGI which blocks
badly on full disk and that's chewing up all their children.
Or perhaps the lockfile can't be locked on a full disk? That'd be cute.
Dean
On Thu, 6 Nov 1997, sameer wrote:
> > Erm.... exactly what are we supposed to do here?
> >
> > rm -Rf / ?
>
> stop logging?
>
> --
> Sameer Parekh Voice: 510-986-8770
> President FAX: 510-986-8777
> C2Net
> http://www.c2.net/ sameer@c2.net
>