[OpenAFS] Re: Server down, other server ready to serve its data: vos delentry question

Jeff Blaine jblaine@kickflop.net
Tue, 01 Aug 2006 11:32:32 -0400

[ Thanks Esther ]

Let's assume for the sake of this discussion that the
old server (downserver) cannot be brought up in order
to syncvldb against the empty /vicep mount points.

Let's also assume that OpenAFS is not being used, but
IBM AFS is being used.  I know this is the openafs-info
list and all, but I'm sure many of you have legacy
experience with IBM/Transarc AFS.  That is, 'vos changeloc'
is not available.

My concern is the wording of the 'vos delentry' instructions:

     "A specified volume can be any of 3 types (RW, RO, or
      backup), but the ENTIRE ENTRY is removed no matter
      what type is provided."

Does that mean if I 'vos delentry -server downserver' and,
while processing the VLDB for things to delete, it will
delete ALL replica site information for a RO that was on


Jeff Blaine wrote:
> Hypothetical Scenario:
>     AFS fileserver 'downserver' is down and unbootable.
>     It served both RW and RO volumes.  A list of these
>     volumes is attainable via vos listvldb.  Let's assume
>     'downserver' is going to have to be down for a week.
>     AFS fileserver 'up' is up.
>     AFS fileserver 'up' has 'downserver's old /vice
>     partitions mounted (they are no longer physically
>     connected to 'downserver').
>     fs has been restarted on 'up'.
> Questions:
> I can't easily test these.
> Obviously a 'vos syncvldb up' needs to happen at some point.
> 1.  Is it necessary (and wise) to use 'vos delentry -server
>     downserver' beforehand to avoid VLDB issues?
> 2.  OR... will 'vos syncvldb downserver' then 'vos syncvldb
>     up' cause the dead VLDB entries to clear out due to the
>     server being inaccessible?