Description:
------------
Runing version 3.0.4 beta
Using memcache session save handler. Prefork apache.

We had an instance where after adding additional virtual host entries (with inidividual log files) after an apache restart CPU hit 100% for each running apache process. 

After running gdb, found a backtrace was always stuck in mmc_pool_select() or mmc_pool_run(). I can't quite remember which - unfortunately I did not save the backtrace and currently failing to reproduce in our dev. environment. I found this blog that has the exact same symptoms: http://blog.roaldseth.net/2009/08/debugging-peclmemcache.html which provides analysis of problem area.

Although this is ultimately a configuration problem (lack of file handles), it would be good if memcache could catch this and not use 100% cpu, which brings down the server.



Reproduce code:
---------------
Unable to reproduce at the moment, though I suspect ulimit -n for the apache process + combination of memcache seession handler will do it.

-- memcache.ini --
extension=memcache.so
memcache.protocol=ascii
session.save_handler=memcache
session.save_path="tcp://XXXX:11211?persistent=1&weight=1&timeout=1&retry_interval=15,tcp://XXXX:11211?persistent=1&
weight=1&timeout=1&retry_interval=15,tcp://XXXX:11211?persistent=1&weight=1&timeout=1&retry_interval=15,tcp://XXXX:1
1211?persistent=1&weight=1&timeout=1&retry_interval=15"


<?php

session_start();

Expected result:
----------------
An error about memcache not being able to obtain a file handle, not high cpu.

Actual result:
--------------
100% cpu for each httpd process doing a session_start()