I am happy/relieved/thankful/whatever to report that our problem
(loss of client machines on ACL's access to those things they were
on ACL's for) is SOLVED!

I received an excellent analysis of what the problem might be from 
Dimitris Varotsis at CMU, who also forwarded that on to Transarc;
Further poking and prodding of the system seemed to follow Dimitris'
model. Transarc turned out a new fileserver binary in exteremely short
order, and I finally got to testing it properly this morning. It works!

Simply put, clients accessing during server outages such as the weekly 
restart, were getting confused. Afterwards, they were staying confused 
until the next restart or a reboot of themselves. The new fileserver cures 
the problem nicely. 

Problems in debugging this were largely due to our being in the first stages
of transferring to AFS - both inexperience, and (initally) very light usage
of AFS making the problem hard to replicate and find a pattern for. This
last Sunday (restarts are Sunday morning here, as per default) the problem
was very repeatable due to higher use of the system.

(AFS 3.2, IBM AIX 3.2)

Thanks to Dimitris, Kathy, and the Development folks doing the coding...

-- 
-Lawrence Smith, MSC Computing Staff - Cornell Univ., Ithaca, NY
 [EMAIL PROTECTED] [EMAIL PROTECTED] (607)255-6064
-Cats, Coffee, Chocolate... Vices to live by. 

Reply via email to