On Sun, Nov 21, 1999 at 12:11:38PM -0500, Thomas Valentino Crimi wrote:
> Excerpts from mail: 21-Nov-99 coda test server by Ben_at_one.net 
> > I just compiled and installed the coda 5.3.3 tarball under Debian Linux
> > (potato). The coda kernel module is loaded. When I point the client at
> > testserver.cs.cmu.edu it "Cannot find RootVolume".
> 
>  One oddity that happened to me when I tried to use
> "testserver.coda.cs.cmu.edu" is that I could ping and nslookup, but
> venus still couldn't find the machine.  I put it in my /etc/hosts file
> and all worked fine.. I'm still pretty baffled by it, BTW.  Only venus
> couldn't find the machine and when I used gdb to see what venus was
> trying to gethostbyname.. it was sure enough correct.

That is possibly related to glibc mismatches. I'm compiling the rpms on
a RH5.2 machine which is using glibc-2.0. RH6.0/6.1 are using glibc-2.1,
but do have a 2.0-compatible library installed as well. My guess is that
is isn't compatible enough.

Adding the server name to /etc/hosts fixes the problem. This is actually
already advised, as it also allows for starting Coda without any network
connectivity once it has the rootvolume in it's cache.


On another note. The problem connecting to the testserver was because it
was down. There is currently a limit of about a thousand clients. Once
that is hit the server fails on an assertion. I guess this is aggrevated
by the fact that these slots are not recycled, and when people are
coming in through a NAT box their portnumbers jump about once every 5
minutes.

Normally this hits during the week, and I restart the server within a
couple of minutes. I've restarted the server now and it should be up
for another couple of days ;)

Jan