[OpenAFS-devel] Re: Breaking callbacks on unlink

[Russ Allbery]

Andrew Deason <adeason@sinenomine.net> writes:
> Derrick Brashear <shadow@gmail.com> wrote:

>> The problem there, to me, is only when those people wish to participate
>> in the global AFS namespace, which is a second issue here, the "play
>> nice or go home" issue.

> Well, I was talking about idledead stuff here, specifically client
> behavior, which isn't going to screw up another client.

If you abandon and retry operations on a locked vnode and hence tie up
more and more server threads waiting on the same vnode, you will
eventually cause a file server meltdown, which most definitely *does*
screw up other clients.

If you retry abandoned write operations on a read-write volume without
ensuring cache consistency with the server, you will create data
inconsistency issues that can result in file corruption, which most
definitely *does* screw up other clients.

If you restrain idledead to only read-only volumes with replicas to which
it is reasonable to fall back, then I agree, you're not going to screw up
other clients.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>