Coda File System

Re: File appeasr different between connected clients

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Fri, 16 Nov 2001 12:21:38 -0500
On Fri, Nov 16, 2001 at 11:11:54AM +0100, Steffen Neumann wrote:
> This message is mainly because of curiosity,
> and to help finding problems.
> 
> Problem: We have a file created on laptop1:
> 
> 	ls -l /coda/homes/floemker/usr_rhea.txt 
> 	-rw-r--r--   1 floemker nogroup   8310678 Nov 16 10:39 
> 				/coda/homes/floemker/usr_rhea.txt
> 
> which appears on all other clients as 
> 
> 	ls -l /coda/homes/floemker/usr_rhea.txt 
> 	-rw-r--r--   1 floemker nogroup         0 Nov 16 10:31 
> 				/coda/homes/floemker/usr_rhea.txt

Ok, this is typical when the create operation got through, but the
filedata hasn't been sent to the servers yet. laptop1 is should be
write-disconnected with at least one CML entry and possibly a
reintegration conflict on a change that might have occured between the
open (create) and the close (store) operations on the file.

However, the 'lv' information shows that the laptop doesn't have any
pending changes.

Try to do a 'cfs er /coda/homes/floemker/usr_rhea.txt' on laptop2, this
triggers an downcall to flush cached information from the kernel. Then
check if the file shows up.

Second try is to disconnect laptop 2, run 'cfs cs', reconnect and run
'cfs cs' again, this should trigger the client to revalidate the
up-to-date status of all objects in the cache.

If that fails, try to do 'cfs fl /coda/homes/floemker' on laptop2, this
tries to flush all objects from the venus cache that are located in that
subtree. Perhaps we got the wrong data with the 'right' version vector.

And let me know which of the three (if any) solves the problem, as this
would narrow the search for the possible cause.

Jan
Received on 2001-11-16 12:22:02