Problems with remote access

[Adam Lipscombe]

We have a remote CVS server that uses SecureShell to access it via an "ext" 
CVS root. Our development team is experiencing repeated CVS access problems 
with this server. The server runs  RedHat 7.2 and has the latest build of 
the CVS server

When we transfer a large number of files (e.g. cvs -n update -d -P) the CVS 
client process on our local PC's seems to hang after a while.
A check with dumeter reveals that there is no data traffic on the wire, but 
the cvs client process is still active.

The CVS server process also seems to be active at this point (at least there 
is a temporary cvs* directory in /tmp on the server). Killing the PC client 
cvs process manually results in the cvs server correctly cleaning up the 
/tmp directory.


As far as I can tell the problem seems to be related to the volume of files. 
Transferring small numbers of files, or individual files always works. I am 
not sure if the problem is caused by CVS itself, or SecureShell. Our 
developers use a variety of cvs clients and the problem happens with all of 
them, so I don't think its client related.

Even the "long" remote CVS sessions themselves are not that long - probably 
15-20 mins at most for a complete update. The problem is that a "long" 
operation virtually always hangs about 2/3rds of the way through.
 Large files such as a .war file almost never work while updates to a 
smaller group generally do. It may be something to do with the file type 
(text or binary) as the problem most often manifests with binary files.

I don't think the problem is inherent to CVS itself as command-line 
operations on the server always work.
IMO its something to do with the transport between remote client and server. 
I have tried using the "-t" flag on a CVS client but the trace doesn't show 
anything useful. It seems like the client is simply waiting for the server 
to respond, which it does not.



Has anyone seen anything like this before, or got any ideas?


TIA - Adam Lipscombe