[OpenAFS] convertROtoRW when the RW is active

Ken Dreyer ktdreyer@ktdreyer.com
Thu, 7 Mar 2013 21:29:32 -0700

There's a convertROtoRW change up for review that Jeff encouraged me
to take to the mailing list. http://gerrit.openafs.org/8014

Currently, "convertROtoRW" checks the VLDB to see if an RW already
exists, and if it does, it prompts the user to continue. The "-force"
flag answers "yes" to this prompt.

As I understand it, Gerrit 8014 would mean that convertROtoRW refuses
to continue if the RW volume on the fileserver is accessible. Even the
"-force" flag would not cause convertROtoRW to continue in this case.

On the one hand, this change may reduce the number of instances where
users could end up with multiple RW copies. I think that's a good goal
in general. AFS can get wacky when there's "secret" leftover copies
floating around.

On the other hand, it seems to me that if I've already specified
"-force" to override the VLDB RW entry, I would also want "-force" to
override the fileserver RW entry as well. I think the code in 8014
should just skip the check in that case (even if the timeout is only
five seconds) and do the conversion. When I have a big site outage,
and I'm having to do a mass convertROtoRW run, I'm already sweating
bullets over how slow the conversion operations are going, and I'd
rather not have vos try to contact the old RW fileservers when I know
they're down. I think "-force" should mean "I know what I'm doing,
just get new RWs online ASAP, and I'll deal with the consequences

In cases where the user hasn't specified "-force", I think it would be
ok to error out, as the Gerrit patch does. Just my 2c.

- Ken