On Fri, Feb 04, 2000 at 06:48:13PM -0800, Tom Tarka wrote:
> Since I couldn't get venus to connect to the SCM server running on my
> MkLinux machine,
> I thought I'd try to set up a replicated server on an x86 Red Hat
> machine and connect to the replicated server with venus.
> 
> I suceeded in getting venus to connect to the replicated server, but
> now I get:
> 
> (tommy_at_sodom) ~ % clog tommy
> username: tommy
> Password:
> (tommy_at_sodom) ~ % ls /coda/
> ls: /coda/: Resource temporarily unavailable
> (tommy_at_sodom) ~ % ls /coda/

Interesting, clog worked, but /coda is inaccessible. But clog does
ioctls on a file in /coda to pass the keys to venus.

Also looking back at your other reports and the logfiles from the server
I get the feeling you have a server-server conflict in the rootdirectory
of the root-volume.

The EWOULDBLOCK is generally a result of failing resolution. We
uncovered some problems with f.i. resolving a directory in 5.3.4. The
just released 5.3.5 should be a little bit improved in that area, but I
don't know whether it can fix the conflict by itself.

Root volume conflicts are annoyingly hard to repair. You'd have to
create a temporary volume, start a reinitialized client with that volume
as the root, mount the original root volume and use repair to fix it.

    createvol_rep tmp_root E0000100 /vicepa
    venus -init -rtmp_root
    clog admin
    cd /coda
    cfs mkm rootvol coda.root
    repair rootvol /tmp/fix

But then again, maybe there is no conflict. Is the root volume
replicated across servers or only on a single one?.

And the comment about the DSL, is that one of those asymmetric ones? And
are both servers on different sides? That's not really a good setup.
Servers are not as adaptive as Coda clients and really like to sit next
to each other on the same LAN. Also clients consider all server in a
replicated group equivalent, and will readily read data across the DSL
link as from the local server. Writes will always go to all servers.
Except when Coda is in weak-connectivity mode, in which case we only
reintegrate to one server (any one again), and then trigger a
server-server resolution.

Jan