It took a little while to figure this out, and hence documenting here …
Need is to restart JBoss once the artifacts are deployed. I have two different jobs, one for building and deploying artifacts (EAR, in this case) and the other one to restart the server. Former invokes the latter as a part of its post-build actions.
I’ve setup a build step in Hudson that executes a shell script which essentially invokes stop and start operations on the server. The server stops fine, but when I start the server in the background, server process gets killed once the Hudson process is finished. I’ve tried multiple ways of achieving this — tried a couple of JBoss Maven plugins and tried Ant route too suspecting if that was an issue with my Shell script. But the real problem lies with the way Hudson deals with the spawned processes. This behavior is consistent with their design according to Hudson docs.
The suggested workaround for Unix systems is to use something like daemonize. That didn’t work for me. Daemonize works fine but the process is still being killed by the Hudson. [Side note: daemonize is a neat tool, glad that I stumbled on it. Need it for some other purposes].
So how was this resolved? Searching the bug tracker, found the exact issue that I mentioned here. The solution/workaround mentioned there works like a charm. Here is what it says —
set the environment variable BUILD_ID to something like ‘dontKillMe’ in the
process that should stay alive.Hudson looks for that environment variable when cleaning up stray processes.
SetEnv plugin is already installed on my Hudson server, and setting BUILD_ID variable value worked! May be that Hudson could provide an option on the admin UI for the user to indicate not to kill the intentionally spawned processes.
