This server crashes every 4-6 hours. Crash here means it has run out of threads / connections so a call to dbtest.tcl fails.
It is not much of a big deal for me, luckily we run load balancing, but I was wondering if other people have found this behavior with regards to running exec "irritating" and have a solution. People have been talking about using ns_proxy for calling exec and I think that is a good idea, but can someone provide a "step by step" guide, so we can create an ns_proxy pool by default for OpenACS 5.4 and use that proxy pool to run all the exec commands through?
Note, it is not only imagemagik, but calls to "top" as done through the request-processor or calls to "find" as done in ]po[ that show the same symptom.
On a related note, I submitted "pools-procs.tcl" which allows you to actually work with the pools in AOLserver 4.5 (as you might have noticed maxconnections and minthreads and so on has no effect if you do not set it up in a pool separately). Additionally I enhanced the monitoring package to give me stats from _ns_stats.* so I could see a little bit more details about the memory usage of the threads and how many threads are running. Now I "only" need a nice way of saving that information before I restart the "crashed" server.