broadcast.go: monitor routines are being terminated and camera not automatically shutting down

[Saxon Milton]

It seems like the monitor routines stop running on occasion. This could be due to either a vidgrind wide crash, or app engine restarting the service. We should firstly investigate the cause of this and then think about a solution. Assuming this is a natural side effect of app engine, we could consider saving livestream details in a datastore variable and then check this with some code on start up of vidgrind. If this saved stream info indicates that there’s a livestream and that it should be monitored, we can spin up a routine to do this.

‌

[Saxon Milton]

Working on this

[Saxon Milton]

We’re going to move away from the routine/background process based system; we don’t need that to achieve a similar sort of monitoring functionality. What can instead be used is the google appengine cron service. The cron will be configured to hit a “/monitor” endpoint that will then look at what broadcasts are active and perform the same sort of checks/tasks. This should be safer and easier to work with when we scale up away from our single instance limitation that’s imposed by the cron scheduler that regrettably adopts a similar background process architecture.

[Saxon Milton]

Merged in fix-monitor-routines (pull request #193)

broadcast.go & broadcast/youtube.go: moving away from monitoring routine to cron initiated checks

resolves issue #181

Approved-by: Russell Stanley

→ <<cset 5e3c4bb15ba1>>