running /etc/init.d/mongrel_cluster start does not clear a previously generated pid. It appears to assume that all pids are in /var/run/mongrel_cluster and forces a mkdir on start. However, configure files in /etc/mongrel_cluster contain an address to the pid file, which is often different (especially on Rails). Reproducible: Always Steps to Reproduce: 1.Create cluster configuration file with mongrel_rails and link to /etc/mongrel_cluster. 2. Start mongrel_cluster with /etc/init.d/ 3. Stop and then restart. Mongrel fails to start because of existing pids Actual Results: Mongrel fails Expected Results: Mongrel starts as expected By adding the option "--clean" to the mongrel_cluster_ctl call in the "start" action of the init.d script, the problem is fixed. Could also be added to "stop"?
I've just had a (very late) look at this bug. Unfortunately I need more information to be able to solve it, since I can't follow you reproduction steps. Please try to be more explicit on what these steps are, specifically what command line options to take. Given that we don't ship an init.d file ourselves I expect this to be an upstream problem, but I can't be certain without understanding what is going on here.
Created attachment 241925 [details] run script for mongrel clusters
Comment on attachment 241925 [details] run script for mongrel clusters It was a long time ago, so long I have forgotten the details. I co recall getting a run script from somewhere and it gave me no end of problems. Attached is the one I have written and it seems to do the job.
I just added this init.d script to the mongrel_cluster ebuild. Apologies for the huge delays here, but I guess the confusion was that it seemed this script was already around while we should have been treating it as a new addition.