[OpenAFS] no quorom elected again

[Ted Anderson]

On 4/30/2004 07:16, J S wrote:
> I get the "no quorom elected" error when trying to create or move a 
> volume on a new partition vicpe, I've just created.

This means that the vlserver is running, but not happy with the 
condition of its peers on other servers.

> # rxdebug -servers localhost -version
> Trying 127.0.0.1 (port 7000):
> AFS version: Base configuration afs3.5 3.60

This is an incredibly ancient file server!  Transarc AFS version 3.5 
(3.60 is the configuration number).  I am pretty sure that AFS 3.6 was 
the basis for OpenAFS, so this server binary has to be 5 years old.  Can 
that be right?

On the other hand there isn't any reason to think that no quorom 
problems are caused by this.  It is just amazing.

 > I've checked the config files and they look OK too. Is there anything
 > else I can try?

The key information would be the contents of your server's CellServDB so 
you can tell what database servers you have in your cell.  Each of these 
servers needs the same CellServDB (the server side file in /usr/afs/etc) 
so that all database servers know about the same set of peers.  Check 
that all those servers are running vlservers.

If all that is working, then you should run udebug on each server 
specifying the port of the vlserver.  Google[1] reminds me that this is 
port 7003.  Interpreting the udebug output is pretty arcane, but there 
are some web pages that explain it.  One easy thing to check for is 
clock skew.  The vlservers need to have the same clocks within a few 
seconds; those machines should all be running ntp[2].

Ted Anderson

[1] http://www.openafs.org/pages/doc/AdminReference/auarf237.htm
[2] http://www.ntp.org/downloads.html