100% cpu utilization of Solaris x86

[Jeremy Allison]

On Thu, Nov 08, 2001 at 06:13:00PM +0000, Neil Hoggarth wrote:
> On Thu, 8 Nov 2001, Jeremy Allison wrote:
> 
> > If that's the case I *DESPARATELY* need a full level 10 debug from
> > startup to "logic error" from you.
> 
> Is tracing individual smbds (logging by seperately by pid, or netbios
> name of client, or whatever) going to be any use, or are you looking for
> one monolithic log.smbd file covering all activites of all clients?
>
> This server has in excess of 100 active clients at any one time on a
> typical day; to be honest I don't know whether I can cope with the
> performance hit (or the shear amount of bytes!) that running with log
> level=10 for any prolonged period will produce. I'm keen to help if I
> can, though.

I need a level 10 log from one client connecting to one server, where
the smbd eventually aborts with a "logic error". There must be something
in the core smbd logic that can get one client to make mistakes in it's
own oplock records in order to get one of these messages, I just can't
see it from code study. Running with seperate logs per client is what's
needed. If you can find an application or client that can reliably
reproduce this, then run that smbd at level 10 from connection to
exit.

> What are your prefered logging options (the smb.conf globals "debug
> <whatever> = {yes|no}" and similar parameters)?

Don't care. I sed out timestamps when reading (and they're occasionally useful
in looking at a timing dependent bug).

Thanks,

	Jeremy.