On Fri, Aug 17, 2001 at 05:04:14PM +1000, Brett Randall wrote:
> Hi All. I am having problems with coda. Please forgive me for the
> length of this e-mail. These are the logs of what I am doing, debug
> level set to 1000.

It looks like the VRDB/VLDB files are not reaching the second server.
Checking the time the getvolumelist is done on both machines, there
seems to be a 2-3 minute clockskew which might be causing this. The
VLDB/VRDB files are kept in sync by comparing the timestamps on the
files and the replica gets confused by the clockskew although updateclnt
is trying to force the same mtime after receiving the file.

These 2 log entries are both from after the creation of the new volume.
[SCM SrvLog]     16:55:47 GetVolumeList returns 0
[Replica SrvLog] 16:53:22 GetVolumeList returns 0

The short term solution is to 'touch' /vice/db/VRDB and /vice/db/VLDB on
the SCM, which should trigger the updateclnt on the Replica to grab the
new versions of these files.

The Replica SrvLog should then show something like
[Replica SrvLog] 16:53:12 New Data Base received

The longer term solution is to bring the clocks back in sync, 2-3
seconds difference would be perfect. The best solution would be to use
xntpd which can bring the difference down to a few milliseconds.

The best solution would ofcourse be to fix updateclnt/updatesrv to care
less about clockskew.

Jan