sourceware.org's web server is not responding

sourceware.org's web server is not responding

[Jim Meyering]

Hi Alasdair,

[now, I see that it's not specific to LVM2 or your commit/log hook]

I've just committed a tiny LVM2 typo fix like this:

    $ cvs commit -F .msg doc/testing.txt
    Checking in doc/testing.txt;
    /cvs/lvm2/LVM2/doc/testing.txt,v  <--  testing.txt
    new revision: 1.2; previous revision: 1.1
    done

    500 read timeout from http://sourceware.org/cgi-bin/cvsweb.cgi/LVM2/doc/testing.txt.diff?cvsroot=lvm2&r1=1.1&r2=1.2

Note the final diagnostic.
Between the commit and the diagnostic there was a long delay.

It didn't affect the commit, since your script is run only on the
cvs post-commit log hook, of course, but it is disconcerting to see
a "cvs commit" appear to hang.

I've just dug a little depeer and see that sourceware's web
server is not responding:

    sourceware$ nc -w2 localhost 80
    [Exit 1]

I can't help with that, but overseers (Cc'd) surely can.

Jim

P.S., Once lvm switches to git, it'll be trivial to include the diff output
in the email-to-list message without having to go through a web server.

Re: sourceware.org's web server is not responding

[Jonathan Larmour]

Jim Meyering wrote:
> 
> I've just dug a little depeer and see that sourceware's web
> server is not responding:
> 
>     sourceware$ nc -w2 localhost 80
>     [Exit 1]
> 
> I can't help with that, but overseers (Cc'd) surely can.

I couldn't see anything that screamed what the problem was, but I 
restarted httpd and it seems to be better.

Overseers: in case it helps, I stored "lsof -i" at the time of the problem 
to /tmp/lsofi. Load was normal. No obvious problems in the httpd logs 
except lots of these sorts of errors in error_log:

[Tue Sep 25 10:27:20 2007] [warn] FastCGI: server "/wiki/moin.fcg" has 
failed to remain running for 30 seconds given 3 attempts, its restart 
interval has been backed off to 600 seconds

This seems like it could be a symptom rather than a cause to me though.

On startup, it rapidly reported:
[Tue Sep 25 10:27:37 2007] [error] server reached MaxClients setting, 
consider raising the MaxClients setting

which isn't so good. Again symptom rather than cause - the question is 
what's making processes (presumably wiki processes?) last so long. The 
logs don't seem to show higher than normal wiki activity.

Also, two other errors were reported:
[Tue Sep 25 10:27:25 2007] [error] FastCGI: access for server (uid -1, gid 
-1) failed: write not allowed
[Tue Sep 25 10:27:25 2007] [error] FastCGI: can't create dynamic directory 
"/etc/httpd/logs/fastcgi/dynamic": access for server (uid -1, gid -1) 
failed: write not allowed

That directory does exist, owned by apache.

Jifl
-- 
------["The best things in life aren't things."]------      Opinions==mine