MsgFlo coordinator didn't stop participant when clearing and re-sending graph

[bergie]

Viewing MsgFlo in Flowhub live mode, switching to edit mode: as result we have two running instances of msgflo-register-foreign with same role:

├─702 node ./node_modules/.bin/msgflo-register-foreign participants/panel.yml --role weltenbaulab --forever=true
├─828 node ./node_modules/.bin/msgflo-register-foreign participants/panel.yml --role weltenbaulab --forever=true

[jonnor]

How reproducable is this? How to reproduce?

[bergie]

@jonnor my gut feel (on Linux), this is quite random ... sometimes all participants stop, and sometimes some of them are left running.

To reproduce you'd have to start the coordinator (which then starts some participants), stop it, start again a few times and see if there are any lingering processes.

I'm seeing this also when running the MsgFlo test suite locally. If I just npm test, I seem to have about 50% chance that there are some Node.js processes left around after the test suite finishes.

[jonnor]

The msgflo test suite does even attempt to clean after itself. But we don't double fork, so I'm suprised that the processes would survive...

Anyways, next time it happens, can you provide also pstree -s -p $PID - where $PID is the process id of one of the participants