search

uaktags / explorer

An open source block explorer
BSD 3-Clause "New" or "Revised" License
6 stars 8 forks source link

Cluster - Need a way to monitor and restart/re-fork workers #25

Closed uaktags closed 5 years ago

uaktags commented 5 years ago

So an issue that I'm finding is that there's occasionally CONNREFUSED happening from the Daemon while the workers are doing their thing. After this error is thrown, the worker is stalled (#24). So we need a way for the workers to tell the master they're still working, and if they're not (but still alive) then the Master needs to kill them and start a new one in its place.

The problem is that cluster.workers[].process.env isn't a thing; You can't restart a worker (no cluster.workers[].restart, etc. You're basically just stuck with a dead process. This almost certainly has to do with the amount of communication we're throwing at the Daemon, but we really need a way to address this.

uaktags commented 5 years ago

So monitoring and creating new workers now work.