Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: CVS Problem (again) ,Slightly lesser old code, but old still [was Re: Console problem ,older code]



On Wed, Nov 18, 2020 at 10:41:31PM -0800, Germain Le Chapelain wrote:
> client_loop: send disconnect: Broken pipe
> cvs [checkout aborted]: end of file from server (conslut above messages if any)

The second line refers to the ssh error in the first line, which is trying to
tell you that "somehow" the tcp connection to the server broke (at the network
layer).

This could happen for various reason. Most typical example I see in local
testing: I ssh into some test machine, start a test, the test makes the
machine panic + reboot - and the client ssh shows the same error you saw
(after some time).

Now this is certainly not the cause for your case (unless you used a local
cvs mirror that crashed), but other possible reasons are plenty. One I have
seen is:

 - you are connected over a NAT gateway/firewall with too low NAT state
   timeouts
 - you cvs update to a quite slow local disk, which causes long pauses
   in the ssh traffic while cvs stats local files
 - when cvs starts to communicate again, the NAT timeout has happened and the
   gateway blocks all traffic

IIRC there were some ssh "keepalive" options to prevent the timeout.

You could use tcpdump -w and wireshark to try to analyze the network issue.

Martin


Home | Main Index | Thread Index | Old Index