Anybody had success with this approach?
TL;DR: I want to push out an ‘exploded’ JRuby-enabled webapp/context
to a Tomcat (6), and am just wondering what the performance etc.
tradeoffs
are vs. a plain WAR based deploy.
background: I’m going to deploy graylog2-web-interface as a JRuby
app to tomcat6 (this works well by the way I’ve done it before).
parts of the functionality of graylog2 is configuring ‘streams’ you can
subscribe to. Stream subscriptions get processed by a rake task run from
cron.
[NB: torquebox might be a better fit for this longer term but I need
something
up and running pretty quickly and I know tomcats quirks fairly well].
so my plan is to run the rake tasks via cron+jruby. I’ve found a way to
do this
against the unpacked WAR
(
http://blog.jonbettcher.com/2010/11/invoking-console-from-deployed-jruby.html?m=1
)
but I’m managing this whole stack with Puppet, and it’s easier to edit
an exploded WAR
than ensure the various settings in the WAR tally up with the rest of
the stack
(elasticsearch, mongo, etc.).
So I’m thinking: warble up the graylog app, expand that on the puppet
server,
template the various config files and push the whole thing down to the
tomcat server
configured to autoload changes in the appropriate context.
Anyone else doing this or something similar? Cheers.