httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Tao <t...@gate.sinica.edu.tw>
Subject Re: leaking fd's
Date Wed, 21 Jun 1995 11:17:06 GMT
On Tue, 20 Jun 1995, Brian Behlendorf wrote:
> 
> It's crept up steadily throughout the day here, from in the low-20's to 
> mid-50's now.  

    That's per child or in total (and with how many children?)  I get
5 or 6 per httpd child, plus whatever else belongs to the CGI, but for
some reason, that number held steady last night when I ran the test
again.  Still trying to figure out what I did differently this time.

    I still have the problem with a lot of zombie children.  I thought
perhaps setting MaxRequestsPerChild to 1 might help, but it doesn't...
after only ten minutes (about 6000 requests), 47 of the 50 children
are in a zombie state.  SIGHUP/STOPping the parent zaps only the
zombies, but leaves the three remaining httpd processes.  Turning
MxRequestsPerChild back up to 30 and running for ten hours leaves 5
out of 50 children still running.  Any idea why the parent does not
restart a child process?  The benchmark includes artificial
"connection drops", which may be causing this... I'll try taking that
out to see what happens.
-- 
Brian ("Though this be madness, yet there is method in't") Tao
taob@gate.sinica.edu.tw <-- work ........ play --> taob@io.org


Mime
View raw message