Coda File System

Re: client died, now what?

From: <jaharkes_at_cs.cmu.edu>
Date: Mon, 26 Oct 1998 20:02:32 -0500
> I was writing a bunch of stuff out to the coda directory, and the client
> (linux 2.1.123, included coda.o, coda 4.6.6) died.  Now, when I start venus, I
> can see the files that were being copied into /coda, but none of the stuff on
> the remote side. Eventually, I get something like this:

This is the first time that I've seen venus crash in the mariner (codacon)
logging routines, and I'm not sure how reinitializing could help. How many 
codacon's are running, did you exit a codacon process? It looks like the
used iterator could result in a bad pointer dereference when dead
connections are destroyed.

It is even possible it can continue when you restart venus, since it looks
like it wanted to print that some part of the reintegration went successful.
(changes are reintegrated in `blobs' of 100, this makes it easier to pick up 
from where it crashed the last time)

If you want to reinitialize, but don't want to lose you disconnected changes.
It is likely that you have not reintegrated all changes. However, there should
be an up to date backup in /usr/coda/spool/<uid>. This directory contains
a tarball with your modified files. If there is not a recent version, it
is possible to force venus to checkpoint the current state (when running)
using the cfs command:
  `cfs checkpointml /path/to/directory'.

succes,
	Jan
Received on 1998-10-26 20:05:59