xinetd problems disguised as cvs problems

Problem

You are running cvs against a pserver. You get:

cvs [update aborted]: reading from server: Connection reset by peer
cvs [update aborted]: end of file from server (consult above messages if any)

Your colleagues have no problems (or only some of them do). Your sysadmin, who doesn't run CVS very often, has NO problems.

Solution

xinetd is crapping on you. If you look in /var/log/cvspserver on the cvs server, you will find entries like:

07/9/12@17:33:09: FAIL: cvspserver service_limit from=$YOUR_IP

What is happening is that you are running up against either the concurrency limits or the connections-per-section limits set in xinetd. To fix, make the following changes in /etc/xinetd.conf

• increase the instances parameter to something higher than 60 (for example, 2000)
• increase the cps parameters to something higher, and lower, than the defaults (like 100 2)

Discussion

The cps parameter controls two things:

• the maximum number of connections per second; and
• the amount of time to take the service out of service should that previous limit be exceeded.