cfservd 2.1.10 trouble?

[Tod Oace]

Is anyone else having trouble with 2.1.10's cfservd? I've been getting 
stale cfservd processes on my cfengine servers ever since I upgraded 
from 2.1.3. The stale processes build up to MaxConnections and then the 
fun begins (lots of cfengine email).

I also found some run-away cfagent processes which weren't helping 
things any. I'm not sure if the two problems are directly related or 
not. I strace'd the processes and they all do this forever:
--- SIGPIPE (Broken pipe) ---
write(1, "cfengine:nwlxsos01: Received sig"..., 259) = -1 EPIPE (Broken 
pipe)
sigreturn()                             = ? (mask now [])

I do have cfagent restart cfservd if it finds it not running. I used to 
use "useshell=false" and then took that out for some reason during the 
upgrade. I've put that back in now.


When I try strace'ing the stale cfservd processes I just get a shell 
prompt back. Which seems odd.

I started cfservd in debug mode but I'm still trying to match up debug 
output with processes. It's tricky without timestamps and pids. "ps 
alxww | fgrep cfservd" shows lots of processes sleeping in "wait_f" and 
"tcp_da".

Anyway, thought I'd check and see if anyone else is seeing anything 
similar before I continue troubleshooting.

--
Tod Oace, Intel Corporation <tod@intel.com>