On Mon, Sep 13, 1999 at 11:44:35PM -0400, squeegy_at_squeegy.org wrote:
> 22:56:24 VAttachVolumeById: vol 1000002 (s.src.0) attached and online
> 22:56:24 create: volume 1000002 (s.src.0) created

Looks good, the volume is correctly built, and attached on the first server.

> 22:57:26 VAttachVolumeById: vol 2000001 (s.src.1) attached and online
> 22:57:26 create: volume 2000001 (s.src.1) created

Same here. Looks perfect.

> If I do an ls -l in /coda this is what I get:
> 
> (chiodi_at_iceman:/coda)--> ls -l
> total 1
> lrw-r--r--   1 chiodi   nogroup         7 Sep 13 22:59 src -> #s.src
...
> the following is the rest of the SCM's SrvLog:
> 22:58:23 vrent::GetVolumeInfo: no hostaddr for volume (33554433)
> 22:58:23 vrent::GetVolumeInfo: no hostaddr for volume (33554433)
> 22:58:32 vrent::GetVolumeInfo: no hostaddr for volume (33554433)
> 22:58:32 vrent::GetVolumeInfo: no hostaddr for volume (33554433)

The SCM?? that's the one on which the volume lookup databases are
created in the first place. The lookup is for a volume with number
2000001 (hex(33554433)), which is the volume on the second server.

And looking at the lookup function, it failed to find a host with hostid
#2 in the list of host addresses.

This list is constructed by reading /vice/db/servers. And it looks like
this file is _only_ read right after a server is started. So you
probably only have to restart the codasrv on the SCM, and everything
will work.

If restarting doesn't help, the volume replication database is bad.

- What is the contents of /vice/vol/AllVolumes on the SCM
- Similarily, what is in /vice/vol/VRList on the SCM

Those files are used as a basis to contruct the volume databases.

Also, the clients sometimes seem to cache volume data too long, in a way
it is a bit like trying to unmount a filesystem while some process still
has hold on an inode. Restarting your client with '-init' should clear
out any bad memories :)

On Tue, Sep 14, 1999 at 07:14:19AM -0400, squeegy_at_squeegy.org wrote:
> Clement,
> 
> Mine looked like yours.  I changed to the one that looks like Jan's
> and still get the same error.  

Mine was wrong, the code here clearly does "%s%d", so that is
"servername serverid", and not "serverid servername" as I wrote earlier.


Jan