OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Subject: Re: rpc.statd
From: Anton Opperman (aoppermafaritec.co.za)
Date: Mon May 15 2000 - 05:21:53 CDT

Olaf Kirch <okircaldera.de> wrote:
> On Fri, Apr 21, 2000 at 10:14:17PM +0100, Chris Evans wrote:
> > What's the purpose of this beast (rpc.statd) in life? Where does it fit
> > into the nfs server framework? What talks to it and why?

1st, wasn't in the office for a while, thus late responce ...

> statd implements the Network Status Monitor protocol (sort of a misnomer,
> it's really a notification service). It is used by the NFS file locking
> service.
>
> It was introduced because NFS is stateless, but file locks are inherently
> stateful. So the problem arises how to deal with locks if either the
> server or a client crashes. In that case, the crashed box, when it
> comes up again, sends out notifications via NSM to statd's running on
> its NFS clients/servers[*]. That statd has been configured to call back
> the local NFS file locking daemon (lockd) so that it can release stale
> locks and/or reclaim previously held locks.

  hmm ... very interesting, and probably acurate, however i haven't
  investigated it thouroughly.

> It's a bletcherous protocol. I have long tinkered with the idea of

  yeah ... on the money. One other interesting "use" of this deamon is
  that it keeps counters on system performance. In order to do this, it
  probably does require root access ... but i beleive that it could be
  reworked not to require this priv'. On old Sun systems, in the SunOS
  days, a system tool (perfmon) that talked to the statd was shipped
  with the OS (I don't have access to newer systems to verify if this
  is still the case).

  Since, I have developed a new performance tool to tap this resource. The
  nice thing about this is that these counters are maintained anyway, thus
  monitoring a system can be achieved with very low overall cost. Also, it
  makes remote (and centralized) monitoring possible, even without any userid
  on the system. With one collection agent, you could even monitor accross
  different OS and architecture boundaries. 

--
  Anton Opperman
  aoppermanfaritec.co.za