I migrated wiki.rubyonrails.org to Phusion Passenger and Ruby
Enterprise Edition about 2-3 weeks ago. Before, the wiki's Mongrels
were crashing at a regular rate, for unknown reasons. We figured that
Phusion Passenger's ability to automatically restart crashed processes
should increase uptime a lot.
It turns out that the wiki has memory usage problems and is constantly
being bombarded by (spam?) bots. Since the migration, I've been
continuously tweaking Phusion Passenger and the wiki's configuration.
For example I've added an option to Phusion Passenger that would
restart the wiki if it used too much memory. Last week or so, Michael
Koziarski and I have finally identified the memory problems. Today the
last memory problem (that we're aware of) has been fixed.
I'm wondering what people think about the wiki's uptime as of late. In
the past few weeks I've restarted Apache a few times in order to
upgrade Phusion Passenger, but the downtime of each such restart was
only a few seconds. In my opinion the wiki has been pretty solid
lately. In the past 2 weeks I haven't seen it go down at all. But a
few days ago someone edited the front page and added the text "It
would be nice if uptime were of some importance to this wiki". So this
made me wonder whether I'm wrong and that the wiki is still unstable.
Comments would be much appreciated.