Re: FREE_ALL locks

New Message Reply About this list Date view Thread view Subject view Author view Attachment view

From: Vern Paxson (vern@ee.lbl.gov)
Date: 08/20/98-02:09:29 AM Z


Message-Id: <199808200709.AAA15706@daffy.ee.lbl.gov>
Subject: Re: FREE_ALL locks
Date: Thu, 20 Aug 1998 00:09:29 PDT
From: Vern Paxson <vern@ee.lbl.gov>

> Of course, the server needs to have a away of probing the connection to see if
> it is still good (to guard from the scenario of a client crashing without
> sending a disconnect) in situations when a conflicting lock. Callbacks through
> firewalls don't work. TCP keep alives could be used. It is also possible to
> have bi-directional RPC on the same TCP connection ...

I agree with you that this latter is better than TCP keepalives (except
the security/firewall issues need some careful thought).  The problem with
keepalives is that they're at a layer that's hidden from the NFS server,
so the server has only crude control over the timing of the keepalives,
and is relying on the underlying TCP implementing an optional feature
(keepalives are not part of the TCP standard).

		Vern


New Message Reply About this list Date view Thread view Subject view Author view Attachment view

This archive was generated by hypermail 2.1.2 : 03/04/05-01:46:12 AM Z CST