So I send a note to support and got a fairly useful reply back. There’s a pointer to an article that seems related. The note also included some interesting info on how monitoring works. However, looking at my current processes they appear to be far below the numbers given in the article, so I’ve asked for some more details on the settings that DreamHost is using. I also suspect that the reply is using the term ‘zombie’ not in the technical sense, since I’ve never seen a true zombie ruby process and the article doesn’t mention that as an issue.
-Nathan
Our logs show that your processes are triggering ProcWatch. Im certain
that you have zombie ruby processes guzzling memory, which is causing
your processes to get killed. I can suggest that you setup a cronjob to
periodically kill these zombie processes. Here’s an article that I found
regarding this topic
http://www.rubyonrailscamp.com/MaxWiki+Trials+and+Tribulations. If you
have any additional questions, please let us know.
2006-12-18 15:33:40 kill 17072 velosa ruby "Ruby
process puts user over the ruby memory limit (21.5703125 / 200)"
2006-12-18 15:43:23 kill 12731 velosa ruby "Ruby
process puts user over the ruby memory limit (21.89453125 / 200)"
2006-12-18 15:45:57 kill 6020 velosa ruby "Ruby
process puts user over the ruby memory limit (21.6015625 / 200)"
2006-12-18 15:53:38 kill 8759 velosa ruby "Ruby
process puts user over the ruby memory limit (21.8984375 / 200)"
2006-12-18 15:53:38 kill 12396 velosa ruby "Ruby
process puts user over the ruby memory limit (21.56640625 / 200)"
2006-12-18 16:03:52 kill 4832 velosa ruby "Ruby
process puts user over the ruby memory limit (21.90234375 / 200)"
2006-12-18 16:03:52 kill 22182 velosa ruby “Ruby
process puts user over the ruby memory limit (21.5703125 / 200)”