[OpenAFS] Upgrade of server by moving drives

Benjamin Kaduk kaduk@mit.edu
Mon, 25 Feb 2019 20:01:23 -0600

On Mon, Feb 25, 2019 at 06:20:02PM +0100, Mike B. Kerber wrote:
> Hello openafs-specialists!
> We are currently upgrading our hardware and would like to circumvent the
> move of the volumes from old to new servers.
> We would prefer to move the array containing our vicep* partitions from
> the old to the new server and bring that one up.
> Downtime is not the issue here but copying the data is.
> Is there a way to achieve this?

This should be a pretty straightforward thing do to.

If the old server is getting fully decomissioned from the POV of AFS, then
it's probably easiest to copy the 'sysid' file from the old server to the
new server concurrently with moving the storage, so the new server will
register itself in the vldb as being the same server (just with a different
IP address, potentially), and clients should update fairly transparently.
(There may be some scenarios that are improved by leaving *a* fileserver
running on the old IP address but with a different sysid, to tell any stale
clients "go check the vldb again", but I forget the details of that.)

Even without sysid tricks, the fileserver just attaches all partitions
mounted at /vicepXX and serve them, so the only difficulty is to make sure
that the vldb is updated accordingly.  The "vos syncserv" and/or "vos
syncvldb" commands can help with this (but I always have to read the docs
to remember which to use).