[Deploying Sakai] AJP Connections Hanging?

[Matthew Buckett]

We recently had an issue where our Sakai deployment stopped respoding
to requests.
We have a frontend Apache which farms out requests to a pair of tomcat
instances over AJP (all Linux, .

When Sakai appeared to stop responding to requests, the load on all
three boxes was low, netstat would show a large number of AJP
connections between apache and the tomcat instances. After a little
bit of prodding we restarted both tomcat instances and everything came
back to normal. The number of AJP connections dropped back
dramatically.

I don't think this is necessarily a Sakai issue but wondered if anyone
else had come across it.
The apache error logs got very busy with the majority of errors being:

(70007)The timeout specified has expired: ajp_ilink_receive() can't
receive header
(120006)APR does not understand this error code: proxy: read response
failed from 10.1.20.2:8009

These errors seemed to appear over previous days but my current guess
is we finally ran out of AJP conections completely and so all requests
would create an error in the logs.

-- 
  Matthew Buckett