cupsd accumulate CPU

angelb angelb at bugarin.us
Tue Aug 1 09:22:59 PDT 2006


> angelb wrote:
> > ...
> > I've never had any problem on polling with CUPS 1.1.23 on this test
> > system. This crash occurred on CUPS 1.2.1 and now I see it again in
> > 1.2.2.
> >
> > I can live with the minimal CPU spike during polling but the daemon
> > crashing is definitely a problem.
> >
> > Where to from here?
>
> Compile with debug enabled (./configure --enable-debug) and then
> run the cups-polld program from the debugger.  When it crashes,
> get a call stack ("where", "bt", etc. depending on the debugger)
> so we can see where it is crashing and maybe why.

Questions: Is there a code in cupsd such that when a certain event
occurs, it kills or send a SIGTERM to its child processes?

If so, what type of events would that be and how do we debug it?

In CUPS 1.2.1, where I saw the same problem, I ran two cups-polld for
more than 24 hours and both stayed alive. Only when cupsd forks-off a
process to start both cups-polld that at any given time, the two
polld daemon dies after receiving a SIGTERM from its "parent". This
occurred on an AIX and Solaris clients. So, myself and another
colleague are leaning towards cupsd as a possible source of the
problem.

Running cups-polld in debugger will not produce an error since it will
not recieve a SIGTERM from anyone.

Lastly, when cups-polld dies, there is no core file. Or at least I've
not seen one.

Thanks,
Angel




More information about the cups mailing list