Cody created SPARK-20422:
----------------------------
Summary: Worker registration retries should be configurable
Key: SPARK-20422
URL: https://issues.apache.org/jira/browse/SPARK-20422
Project: Spark
Issue Type: Bug
Components: Spark Core
Affects Versions: 2.1.0
Reporter: Cody
Currently in standalone mode, worker registration retries is fixed: 6+10 = 16.
which means if a master is down and there's no available standby masters, in
about 15min, the whole cluster will go down, including all workers, executors,
drivers, which is unacceptable in production. I think a master failure should
only affect resource allocation rather than killing a whole cluster, at least
users should have a choice, which means retries should be configurable.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
