[OpenAFS-devel] odd fileserver/volserver interactions

[Robert Banz]

Hi everybody,

I've been noticing some odd fileserver/volserver interactions over the 
past few days, and I'm wondering if anyone out there has any thoughts.

What seems to be happening, is that during time time when I'm sending 
.backup volumes to tape, I'll get periods when my "idle threads" drop, 
and my "calls waiting for a thread" rises to some numbers significantly 
above '0' (where it usually sits).

There doesn't seem to be an extreme amount (or any really) I/O 
contention in play here, iostat still shows reasonable svc_t's for those 
devices.

These machines are running OpenAFS 1.4.0.

I can cause the 'fileserver' behaviour to return to normal by kill 
-9'ing the volserver, however, I don't think this should be happening at 
all...

Do volserver operations on a volume -- even a .backup volume -- cause a 
lock to be placed on it?  Would this lock in turn cause a fileserver 
service thread to block?  I would assume the fileserver would/should 
simply fail and return a VBUSY when seeing such a lock and continue on 
servicing requests, instead of holding up a thread?

Thoughts anyone?

-rob