[OpenAFS] Re: Questions about multihoming servers

Andrew Deason adeason@sinenomine.net
Tue, 24 Sep 2013 17:35:57 -0500

On Tue, 24 Sep 2013 18:01:08 -0400 (EDT)
stephen@physics.unc.edu wrote:

> - Is there any obvious reason to choose multihoming of a fileserver
> rather than link aggregation, assuming both are supported in a given
> environment?

I don't think so, except multihoming can be easier, since you don't need
to do anything for it. I think multihoming can be worse in some
situations, since if the server is actually down, a client may retry
reaching it for every IP it knows about for the server, so it can take
longer to notice.

> - The Admin Guide on docs.openafs.org indicates that multihomed DB
> servers work.[1]

That link seems to be about fileserver multihoming, but I'm just
skimming it. Regardless:

> Specifically, I'm curious about linux dbservers; clients may contact
> the server via IP A but get an answer from IP B, if it is the default
> interface. Will the clients care? What if the IP replying isn't in the
> client's CellServDB (or DNS) at all? Does the answer change depending
> on client version?

dbserver vs fileserver doesn't seem to matter for this. If a client
initiates an rx connection to a certain IP, it can get a reply from any
other IP and continue; I believe that was deliberately done because of
the potential problems you're talking about. This occurs at a much lower
level than stuff dealing with DNS or CellServDB, etc.

This situation might be different if dbservers initiated new connections
to clients, but they don't do that. Fileservers do that for callback
breaks, but I don't think we currently pay attention to the source IP
for those at all.

Andrew Deason