Re: [Nfsen-discuss] problem with more than 10 sources with the latest snapshot

[Yann B. <yb...@ba...>]

On Wed, 09 Aug 2006, at 16:01, Peter Haag wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hi Yann,
> 
> - -------- Original Message  --------
> From: Yann Berthier <yb...@ba...>
> To: nfs...@li...
> Subject: [Nfsen-discuss] problem with more than 10 sources with the latest	snapshot
> Date: Wed Aug 09 2006 13:53:51 GMT+0200 (CEST)
> 
> >    Hi Peter,
> > 
> >    It seems that i triggered a bug when trying to have more than 10
> >    sources, probably in the pid generation routine: nfcapd logs
> >    "Another collector with pid 6312 is already running, and configured
> >    for '/data/nfsen/profile/live/cat2'" at start time, so the nfcapd
> >    process for this source is not spawned
> > 
> >    Reproducable on another machine, another OS, with 12 sources:
> > 
> >    Shutdown nfcpad: 6:[8912] 11:[8917] 3:[8922] 7:[8927] 9:[8932]
> >    12:[8937] 2:[8942] 8:[8947] 1:[8952] 4:[8957] 10:[no pid file found!]
> >    5:[no pid file found!].
> > 
> >    i have not tried with older releases
> 
> Well it seems, as you really configured the same data directory '/data/nfsen/profile/live/cat2' twice:
> nfcapd recognizes if there is already another process running, using the same data directory. It checks if, the other process is really running, or dead. This is not based on the pid file, but on
> shared memory which nfcapd/nfexpire are using to communicate. ( IPC ).

   sorry, i supposed naively that - ok never mind

> So in your case nfcapd really found this process running. I guess there is a configuration issue, or a process which is still hanging around for some reason. Stop all nfcapd processes and check in the
> ps output, if really all nfcapd collectors stopped. There should be no more shared memory segements allocated.

   I can confirm that no shared memory segments are left behind when i
   stop nfsen, no process, no nothing, nada

   so, same test on a gnu-linux-fedora-whatever-fc4 machine and a
   freebsd-current box. Well, almost: 11 sources there, 12 here

   i have x unique sources declared. I can confirm that after starting
   nfsen, along with some error messages at start time, like pasted in
   my previous mail, or each 5 minutes, like below:

   "Expected file not found:
   /data/nfsen/profiles/live/5//nfcapd.200608091625"

   i have no more than 10 shms on both machines owned by my nfsen user,
   and i have no more than 10 udp sockets for nfsen, and no nfcapd
   processes for the cursed sources (so one for the 11 sources case, and
   2 for the 12 sources case)

   if you have the clue bat ... in the meantime i'll dig around to see
   where i could have messed up something
   
   thanks,
   
      - yann