hozer_at_drgw.net said:
| Write disconnected operation is definitely faster, but still has bugs
| in it.. I've ended up reinitializeing servers twice because of it. ;) 

I haven't seen any good server corruption for a while now... and we've
been discussing/working on a change in the kernel-venus protocol that
should get rid of the last problems that crop up during the local ->
global fid replacement.

| Have you ever checked load on the server? When doing something like
| reintegration or untarring a kernel, the load on my server(s), codasrv
| takes around 40-50% CPU. (this is on a PII-450, and from only one
| active client!)
|   Another strange thing.. The SCM machine seemed to have a
| consistently higher CPU load from codasrv by about 10%.. The servers
| are pretty much identical also. Can anyone think of reasons for this,
| other than possibly one disk being slower?

Yes, I have tried to find out why the servers seem so sluggish, and most
of the time seems to be spent in the kernel trying to flush the buffers
for RVM. I got that from strace'ing a server while untarring the
coda-source tree, measuring the time spent in the system calls.

The higher CPU load on the SCM could be partially fixed by now. All
fetches were done from one server, the last or first in the sorted list
of ip addresses. We activated some code that makes the clients switch
their idea of the primary server once in a while.

Jan