[OpenAFS] Re: Performance problems seem to be coming back

Andrew Deason adeason@sinenomine.net
Fri, 9 Sep 2011 15:42:27 -0500

On Fri, 09 Sep 2011 15:36:16 -0400
Dale Pontius <pontius@btv.ibm.com> wrote:

> That's certainly better than using regular ping, since it's going to
> go through the AFS message queues, etc.  In fact, the coworker next
> door has been having some problems, and we tried this method in
> another window on another client machine right as his script slowed
> down, and saw a 20 second round trip time.

rxdebug (in this method of operation) doesn't work just like "ping"
does. Returning after 20 seconds doesn't mean packets take 20 seconds to
traverse back and forth; it probably means some packets were dropped and
we just retried. I think rxdebug waits for 1 second, and then retries
and waits 2 seconds, then 4 seconds, etc. We probably could provide
something closer to a normal "ping" utility.

Andrew Deason