On Fri, Jun 09, 2000 at 10:53:18AM -0700, Phil Nelson wrote:
> 
> Hi,
> 
>   I'm trying to push coda a bit on these big servers.  I set up 
> a 1G RVM server (20G /vicepa) using private mapping using the
> vice-setup scripts.
> 
>   I then set up a client and started creating volumes.  My next
> test was to untar a 26M tar file.  A couple of things happened:
> 
>   a) tar stopped (venus was doing reintegration) and didn't restart
> 	until I did a ^Z and a fg.
> 
>   b) Part way through the untar the server quit with the following message:
> 
> Assertion failed: vm_inuse->Value(*index) == 0, file "../../../coda-src/resolution/recov_vollog.cc", line 618
> VENUS IS EXITING! Bye!

This one is in AllocViaWrapAround, aka. the server ran out of logspace
on the volume because the client didn't send any (enough?) COP2
messages.

Increase the logsize after restarting the server (volutil setlogparms)

Was the client weakly connected, reintegrating only with one server?

We better first check whether both volume replicas are accessed
correctly, there could be a problem with either the VRDB or the VLDB.

To test the VLDB, mount the undelying replicas in different places.
 (cfs mkm rep0 replica.0 ...)

To test the VRDB, write a file to the replicated volume and then check
the mounted replicas to see if all (both) of them have a correct copy of
the file. Maybe all operations consistently fail on one server.

Jan