Following our discussion on #g-infra, the problem with fastcgi b0rking with untrapped signal 6 can be solved (bug 79557) with a simple fix to ruby-fcgi. Another useful change would be to replace mod_fastcgi with mod_fcgid (bug 79313). mod_fcgid has much better process management. It can spawn new processes in a more controlled way, it can decrease the number of processes when the load goes down and it can recycle processes automatically. I can set up the config files so that all you have to do is edit /etc/conf.d/apache2 to switch between fastcgi implementations. Just s/-D FASTCGI/-D FCGID/ and vice versa in APACHE2_OPTS to switch.
Got kiwi moved over. Going to work on getting all our nodes switched over in the next coming weeks.
Xavier -- What, if anything, remains to be done with this bug?
This will get resolved when I push out the new www node refresh. All the nodes except bluejay (Because its still axkit) are running mod_fcgid.
bug cleaning ping
I'm getting close to getting this fully implemented. I'm working on getting viewcvs moved over to a decidated box, then I'm working on getting the new master web node up (loon) and gradually point wren and kiwi to it. I'd guess an ETA of a month or less. That of course depends on my free time. I basically just need to test the cfengine end of things at home a bit more and distribute it.
Can't think of anything left to do.