www-apache-bugdb mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From David Gilbert <dgilb...@velocet.net>
Subject Re: general/3763: Keepalives hang around, using up all the process slots.
Date Tue, 26 Jan 1999 08:27:29 GMT
>>>>> "marc" == marc  <marc@apache.org> writes:

marc> Synopsis: Keepalives hang around, using up all the process
marc> slots.

marc> State-Changed-Why: As we ask before you submit a bug report,
marc> please try with the base Apache config without extra modules.
marc> It is possible that the threads stuff is messing things up, or
marc> that one of the third party modules is messing things up.

It's very possible.  The only module I can subtract is the python
module, if you really want that.  The other modules (php3, ssl) are
required for server operation.  This doesn't seem to show up under
light load.  We've found a number of additional wierd things about the
server (and I was probing the bug database --- something about the
alarm code being changed --- which may be in play here):

1) If keepalives are on, some small percentage of the keepalives stay
   alive indefinately.

2) If you kill -USR1 the server, the G's never seem to go away.

3) Some small percentage of W state processes start chewing up  100%
   of available CPU while serving an ordinary file (such as a gif or
   jpg) that shouldn't be triggering any modules.  This seems to be
   random files --- it's not reproducable on a specific file.

Would you like me to subtract python?   I have no test bench to run
the problem unless php3 and ssl are in the server.


|David Gilbert, Velocet Communications.       | Two things can only be     |
|Mail:       dgilbert@velocet.net             |  equal if and only if they |
|http://www.velocet.net/~dgilbert             |   are precisely opposite.  |

View raw message