Slave hangs when started before master

[tnachen]

Currently when the slave is brought up with just --slave and default options, if the master is not available or not running yet, the slave will just hang and never attempt to reconnect or retry it's calls. No exception will be thrown either, so it just runs quietly.

[kainoaseto]

+1 This would be fantastic to help make Locust more HA. Allowing the slaves to poll for a master would at least allow the master to be brought up asynchronously and would allow for the master to crash and be restarted without having to restart all the slaves if a problem occurred.

[cgoldberg]

@tnachen

if the master is not available or not running yet, the
slave will just hang and never attempt to reconnect or retry it's calls.

If I start a slave with no master, it sits quiet and polls for the master node. If I start the master node afterwards, slaves seem to connect fine. Are you seeing different?

It is slightly confusing that there no polling status is printed, whereas master reports "Waiting for slaves to be ready" each time it polls.

[cgoldberg]

@kainoaseto

allow for the master to crash and be restarted without having to restart all the slaves

It currently only tries to connect before hatching starts. Reconnecting if master crashes is a more complex case.