Re: cfservd memory/file handle leak?

[Mark Burgess]

I have already spent a lot of time on this but have not found any
possibility for it to happen. So I am stumped. This is therefore not
high on my radar list, but I accept any clues as to the possible cause
and do what I can to find these things.
As you say, it is not easy, esp in a multihreaded program

M

On Thu, 2005-12-22 at 09:03 -0600, Paul Krizak wrote:
> The leak is quite problematic on a large network.  I've taken 
> preventative measures by increasing the number of file descriptors for 
> the cfservd process and restarting it every morning, but 1 year from now 
> when I've got 10,000 clients (or more), this will mean having to restart 
> cfservd midway through a scheduled cfengine run just to keep it from 
> crashing, as the process will try to garner over 20,000 file handles 
> that it refuses to release.
> 
> I fully understand the complexity of locating leaks like this (I've 
> dealt with my own leaky programs), but I sincerely hope that "just 
> restart it once in a while" is not the accepted solution to the problem 
> and that some time is devoted to locating the leak.
> 
> Paul Krizak                         5900 E. Ben White Blvd. MS 625
> Advanced Micro Devices              Austin, TX  78741
> Linux/Unix Systems Engineering      Phone: (512) 602-8775
> Microprocessor Solutions Sector
> 
> 
> Mark Burgess wrote:
> > On Wed, 2005-12-21 at 09:51 -0500, christian pearce wrote:
> > 
> >>What is the best way to track down this leak?
> >>
> >>On 12/21/05, christian pearce <christian.pearce@versatile.com> wrote:
> >>
> >>>Has anyone tried the latest and greatest?  Is it fixd?
> >>>
> >>>USER       PID %CPU %MEM   VSZ  RSS TTY      STAT START   TIME COMMAND
> >>>root      2491  0.3 26.3 1080636 1069048 ?   S    Nov22 129:29 cfservd
> >>>
> >>>I am running 2.1.15 on RedHat.  It has been running for over a month
> >>>now and it has 1 gig of memory consumed.
> > 
> > 
> > 
> > nexus$ ps -elf |head
> >  F S      UID   PID  PPID  C PRI NI     ADDR     SZ    WCHAN    STIME
> > TTY      TIME CMD 8 S     mark  8065  8050  0  50 20        ?
> > 145        ? 09:43:41 pts/3    0:00 grep cfser
> > 
> > 
> > Looking at my own sytsems I can see a slow growth of VM size over 3
> > months, on a busy server, nothing dramatic. I suggest that you randomly
> > restart the daemon to avoid this. That is what we used to do for dns
> > which has a much worse leak. Debugging this is not easy. I have gone
> > over things many times. It could also come from some third party library
> > like openssl or berkeleydb. It is very hard to know. But it is not a
> > serious problem. This is not going to tax anyone's servers.
> > 
> > M
> > 
> > 
> > 
> > _______________________________________________
> > Help-cfengine mailing list
> > Help-cfengine@gnu.org
> > http://lists.gnu.org/mailman/listinfo/help-cfengine
>