Howto monitor cupsd ?

[Nathan Whitehorn]

>
> Hello,
>
> On Nov 7 02:14 Nathan Whitehorn wrote (shortened):
> > > ... the cupsd is sometimes running, but not responding.
> ....
> > I'm having the same problem (CUPS 1.2.2, FreeBSD 6.1)
> > ... printing via IPP, often from windows
> > It occasionally happens (once or twice a day) that the CUPS daemon
> > will stop processing requests. The accept queue (as seen from netstat -aL)
> > will increase to maximum, and nothing further happens until the daemon
> > is killed.
> > There's no obvious correlation with anything, but it's hard to tell,
> > because if LogLevel is set to Debug, the log grows several megabytes
> > a second. On Info, there's nothing obvious.
>
> Why don't you show at least some of those tons of messages
> which there are in your error_log?

Because they aren't interesting. Everything is fine (job completely processed or whatever), and then everything stops. Also, keeping those logs puts a severe performance penalty on the server due to disk accesses.

> Why don't you tell at least some basic info how you
> have set up your CUPS and Windows print systems?

I did, if you read the original post. Users are printing via IPP with HTTP basic authentication routed into a PAM stack. On Windows, they are submitting jobs in postscript, but that's not relevant because the fault is in the scheduler, not the filter code.

When this happens, CPU usage goes to zero (from ~ 15%), and the process goes into a wait state. It appears to be doing a blocking read on something that will never produce data. Due to the single-threaded nature of the scheduler process, this results in the server hanging, permanently.
-Nathan