You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Joe Schaefer <jo...@yahoo.com> on 2010/03/07 19:47:22 UTC

memory leak in 2.3.4-alpha

Unfortunately all the httpd hackers in infrastructure have
wandered off the reservation, but we've been running 2.3.5
for the Apache websites for the past coupla weeks and it
has a rather serious memory leak somewhere.  That's the core
reason why we ran into those gzip encoding issues in February.


If any httpd committer familiar with Solaris 10 wants to take
a look at the server, I'd be happy to arrange access.  Just
hop on #asfinfra on freenode.

Thanks.


      

Re: memory leak in 2.3.4-alpha

Posted by Jeff Trawick <tr...@gmail.com>.
On Sun, Mar 7, 2010 at 6:54 PM, Joe Schaefer <jo...@yahoo.com> wrote:
> Hm, on closer inspection it seems the high memory
> usage we're seeing is common to both 2.3.5 and 2.2.12,
> so it's likely an mod_mbox issue, which is probably infra's
> ball of wax to deal with.

Looking at backtraces from brk() could help.

# /usr/sbin/dtrace -n 'syscall::brk:entry /execname == "httpd"/ {ustack();}'

Re: memory leak in 2.3.4-alpha

Posted by Joe Schaefer <jo...@yahoo.com>.
Hm, on closer inspection it seems the high memory
usage we're seeing is common to both 2.3.5 and 2.2.12,
so it's likely an mod_mbox issue, which is probably infra's
ball of wax to deal with.

If anyone committer wants to help, the invitation is still
open, but this time it's for eos (another Solaris 10 host).

Aurora, the box I was talking about in the message below,
failed to survive a reboot, so the issues we've been dealing
with it in infra may not be the fault of code in httpd.



----- Original Message ----
> From: Joe Schaefer <jo...@yahoo.com>
> To: dev@httpd.apache.org
> Sent: Sun, March 7, 2010 1:47:22 PM
> Subject: memory leak in 2.3.4-alpha
> 
> Unfortunately all the httpd hackers in infrastructure have
> wandered off the reservation, but we've been running 2.3.5
> for the Apache websites for the past coupla weeks and it
> has a rather serious memory leak somewhere.  That's the core
> reason why we ran into those gzip encoding issues in February.
> 
> 
> If any httpd committer familiar with Solaris 10 wants to take
> a look at the server, I'd be happy to arrange access.  Just
> hop on #asfinfra on freenode.
> 
> Thanks.