On Mon, Sep 08, 2008 at 10:50:57PM +0100, Jose Calhariz wrote:
> 
> I have a similar problem in a similar setup.  'vos' commands that
> manipulate VLDB don't finish.  My setup is Two AFSDB servers running Debian
> stable with the lowest IPs + 3 older AFSDB servers.  6 Fileservers,
> some of them in the same machines than the AFSDB servers.
> 
> My own research have show that one of VL Server was restarting with
> signal 6, if I remember well.  After the restart the server it don't
> see more messages like that.
> 
> I can do 'udebug server 7003' for all the servers, the server with the
> lowest IP have the following fragment:
> 
> I am sync site until 57 secs from now (at Mon Sep  8 22:28:18 2008) (5 
> servers)
> Recovery state 1f
> I am currently managing write trans 0.4852
> Sync site's db version is 1220892297.1
> 0 locked pages, 0 of them for write
> There are write locks held
> There is an active write transaction
> Transaction tid is 0.0
> 
> No other AFSDB servers says they have write locks.
> 
> The best way is to restart this server?  As this AFSDB server is a big
> fileserver, I expect that a restart will put almost half my users
> volumes down for 2 hours.  If everything goes OK.
> 
> I seek advice, as in the other thread everything went fine with the
> rebuild of the faulty AFSDB server.
> 
>        José Calhariz
> 
> 

Some vos commandos didn't finished like
vos listaddrs
vos examine root.cell (with tokens active)

vos examine root.cell -noauth (sometimes finished, sometimes didn't)

Some clients couldn't start AFS services.

I have restarted all vlservers with active write transaction doing:
bos restart -server $server -instance vlserver -localauth

This seams to fix all the problems.  For the record if someone else is
in the same situation.

     José Calhariz


-- 
--
"Somente 3 coisas páram no ar:
Helicóptero, beija-flor e Dadá Maravilha"
--Dadá Maravilha

Attachment: signature.asc
Description: Digital signature

Reply via email to