Github user IgorBerman commented on the issue:
https://github.com/apache/spark/pull/20640
@squito so I've run internal workload with blacklisting on, with killing
executors for the framework manually. It didn't have much effect(only once got
blacklisting 1 executor and then after timeout it removed it)
If you have some idea how to test it properly I can try to do it
specifically aiming to this scenario of rejecting offers from mesos master for
blacklisted slave resource.
ps: not strictly connected to this conversation, but as a side note: I've
probably found bug that running in coarse grained mode on mesos +
dyn.allocation on + blacklisting on(not sure if it's relevant) + client mode
and killing manually executors on some slaves makes spark driver to start 2nd
executor on same slave for the same application(which is against constraints of
coarse grained mode?)
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
