dutch-master
Advanced tools
Changelog
v0.1.7
Changelog
v0.1.5
Explicitly stop workers when shutting down, since cluster.disconnect does not handle workers that do not close cleanly.
Reduce timeouts, making sure the individual worker timeout is much shorter than the cluster shutdown timeout, so the cluster won't be killed before the workers have been cleaned up.
Changelog
v0.1.4
Only listen for worker 'listening' event once
Modules like statsd clients that create ephemeral sockets can cause many 'listening' events to occur. Make sure dutch-master only responds to the first one.
NOTE: Workers should ensure that the first 'listening' event that occurs signals their readiness to join the cluster i.e. do not create an ephemeral socket as part of the worker startup before starting to listen for actual connections.