Solaris 8, samba 2.2.8a, high smbd CPU
M A Young
m.a.young at durham.ac.uk
Mon Apr 28 14:39:19 GMT 2003
On Mon, 28 Apr 2003 M.Maclaren at bath.ac.uk wrote:
> >> In these cases, smbd logging shows repeated api_DosPrintQGetInfo: uLevel=2
> >> client calls - as if the client is not getting an answer it appreciates.
> >> - in the cases I've investigated, the queues have been empty,
> >> and it isn't always the same queue.
> >> use client driver = yes
> >> disable spoolss = yes
> >> printing = sysv
> >> printcap name = /opt/packages/samba/lib/smb.printcap
> >> load printers = yes
> >> lpq cache time = 60
> >Yes, this is because you are disabling spoolss, so the
> >clients loop requesting printing status.
> Maybe I'm missing something here, but shouldn't clients default
> to lanman style printing and work without the above looping?
The requests from windows are still there even if you revert to 2.0.10a,
but they seem to have much less CPU impact than on 2.2.8a, maybe because
it can just return an error rather than thinking what the answer should
be. We have just had to revert from 2.2.8a to 2.0.10a (with a backported
printer memory problem fix), because the newer version was much heavier on
cpu load in our 80 odd printer, several hundred windows clients situation,
maybe because of this or related problems.
More information about the samba-technical