Hello,

I decided to build a coda filesystem to hold a source code repository that
will be managed by PRCS.

I installed coda 5.3.19 from /usr/ports on FreeBSD 4.5.  My RVM data area
is a 90M raw partition; my RVM log area is a 20M raw partition on a
separate disk.  My log size is 2M (should I have said 20M?), so vice-setup
reports:
  --- log area: /dev/rdas1a, size 2M.
  --- data area: /dev/rda0s1h, size 90M.
/vicepa is a 1GB filesystem.  There's plenty of space under /usr/coda and
under /vice.  I installed venus on the same machine with a 20MB cache.

My problems came when I tried to check-in the source repository.  This is
an intensive operation (not unlike untarring a large archive), where PRCS
creates about 5000 directories and RCS files, totaling about 30MB.  The
checkin produced lots of coda errors and an unusable, corrupt source
repository -- for instance:

20:32:51 Creating new log for root vnode
20:32:51 VAttachVolumeById: vol 3000002 (prcs.data.0) attached and online
20:32:51 create: volume 3000002 (prcs.data.0) created
[...]
rcs: /coda/prcs/PRCSroot/www/f/46_star_readi,v: No space left on device
WorkerMux: worker read error
[...]
20:56:01 Caution, replicated volume 0x3000002, resolution is turned off.
20:56:31 Reintegrate: prcs.data, 100/518 records, result = SUCCESS


I recall that coda can have a performance problem with intense operations
like this checkin.  I'd like to know if I can tune my configuration so
that it will work.  Can anyone help?

Thanks,

-------------------------------------------
John David Duncan
Systems Administrator
Great Schools, Inc.