On Thu, Nov 15, 2001 at 09:03:16AM +0800, Jeremy Malcolm wrote:
> volumename mountpoint for the four replicated volumes I now immediately
> 
> 08:46:57 Caution, replicated volume 0x1000001, resolution is turned off.

Ok, this is a pretty innocent warning as long as 0x1000001 is a
replicated volume with only one replica. Nice to know that cfs mkm is
still triggering attempts to spool resolution log entries even when
resolution is disabled/unneccesary.

> and when I then do ls in /coda I see this:
> 
> 08:47:53 vrent::GetVolumeInfo: no hostaddr for volume (2000001)
> 08:47:53 vrent::GetVolumeInfo: no hostaddr for volume (2000002)
> 08:47:53 vrent::GetVolumeInfo: no hostaddr for volume (2000003)
> 08:47:53 vrent::GetVolumeInfo: no hostaddr for volume (2000004)

Nice, that's a message that I think Shafeeq added, and this should lead
us pretty closely to the real cause of the problems.

During startup, the server reads /vice/db/servers and creates a mapping
of host identifiers (number between 1 and 254) to ip-addresses by
resolving the hostnames in /vice/db/servers. In this case the server
that logs these messages either never saw and entry for server 2, or the
gethostbyname returned a failure.

So the problem is either an incomplete version of /vice/db/servers on
this machine, or a hostname lookup failure at the time the server is
started.

Jan