Re: Recovering from a crashed file server?

Ken,

On 4/5/2022 6:46 PM, Kenneth Aaker ([email protected]) wrote:
> I've been messing around with this since last Thursday, so be gentle. 
> If this is an inappropriate venue, just tell me.

The openafs-devel mailing list is intended for discussions related to
the software development of OpenAFS. 

The openafs-info mailing list would have been a better forum for posting
a question for community support.


> Anyway, the fundamental question seems to be "Can I use vos
> syncvldb to get the existing volumes for /vicepa, ... to show up again
> in the file server?"  The bit of the file server that died was the
> hard drive that the system was installed on, so I've lost every bit of
> OpenAFS server configuration that I had accumulated...  The /vicep*
> (in particular /vicepa, /vicepb, /vicepc, and /videpd) areas seem to
> be intact (they were on other hard drives). But, I can't get any of
> the volumes from any of the /vicep* directories to be served by the
> file server process. The vos command insists that the VLDB has "Total
> entries: 0" and displays nothing.  But a vos command to list volumes
> insists that there are 40 volumes on the file server?
>
> At this point, I'm nursing scars from attempting to figure out what is
> going on, so I'm a little reluctant to just try a "vos syncvldb" command.
>
> Advice?
>
Each fileserver has associated with it a UUID that is generated the
first time the fileserver process is started.  This UUID is used to
register the fileserver with the (volume) location service creating a
binding between the fileserver, its IPv4 address(es) and the volumes
stored on the fileserver's vice partitions.

"vos listaddr -printuuid -noresolve" will output the UUID for each
registered fileserver.   My guess is that in your cell there will be an
out-of-date fileserver entry for the UUID that used to be associated
with the fileserver and a new entry for the current UUID.  

If this is the case, the volumes will be associated with the old UUID in
the location database.

To synchronize the volume locations with the new fileserver UUID, you
need to execute "vos syncvldb" followed by "vos syncserv".   You can use
the -dryrun switch with each command to see what changes the command
will make to the location database.

http://docs.openafs.org/Reference/1/vos_syncvldb.html

http://docs.openafs.org/Reference/1/vos_syncserv.html

Jeffrey Altman