Defunct (Janitor) process after OFFLINE,ONLINE (or INIT) about ngas

[rtobar]

@szampier thanks for reporting this and sorry for taking so long to respond, but I was on leave.

After trying it out I can also see the same behavior. I was able to at least duplicate the problem with the following:

• Start the server data checker disabled (for simplicity) and started with -autoonline, I see two processes (main and janitor)
• Send OFFLINE command, I see only the main process
• Send ONLINE command, I see the main process plus a defunct one

INIT is just OFFLINE + ONLINE in one go, so it would be the same.

The problem was that the code starting/ending the background tasks didn't clear the Event object before starting the background tasks. These events are what we use to coordinate the shutdown of the background tasks when necessary. This is usually when SIGTERM/SIGINT is sent to the server, but also during the OFFLINE command. When shutting down the background tasks we call event.set() and wait for the background tasks to timely react and finish -- but we never clear the event again, so a new ONLINE command started the background tasks, all of which immediately finish, hence the defunct process (the server would have collected it when shutting itself down).

I'll put a patch in now for this, it should appear in master in not too long.

[rtobar]

Done, a fix for this is now on the master branch.