[OpenAFS] Re: Afs User volume servers in VM's

Renata Maria Dart renata@slac.stanford.edu
Wed, 26 Oct 2011 12:09:17 -0700 (PDT)


>How many clients?

That will vary....sometimes 1000+ but I have seen thread
depletion with just a handful of clients all writing to the
same file or directory.

>How many outstanding RPCs?

Up to ~350 calls waiting when it is at its worst.

>How many threads on the file server?

'/usr/afs/bin/fileserver -L -p 128 -udpsize 262144 -sendsize 262144 -vc 4000 -s 40000 -l 12000 -cb 1500000'

>Do these processes communicate with each other to synchronize their work
>and as such rely on the AFS cache coherency semantics?  If not, could
>these jobs be modified to open the files in a different manner that did
>not enforce those semantics?

The problem is that we get new physicists joining us from time to time.
Our old hands know how to structure their jobs to avoid causing problems
for AFS (and nfs, I might add).  But new users, unfamiliar with AFS,
inadvertently cause problems.  And as Booker mentioned the problems are
often caused by invoking a utility that unbeknownst to them writes
to their home directory.  New users eventually learn, but until they
do, it is a problem.  We have information about best practices posted
as do many of our individual experiments, but new users don't always
get the word.