Github user suryag10 commented on the issue:
https://github.com/apache/spark/pull/22433
I am observing some weird behaviour when i am trying to respond to the
comments. Hence i am adding the resposes to comments as below.
Following are the responses for the comments:
>>The script may be run from a client machine outside a k8s cluster. In
this case, there's not even a pod. I would suggest separating the explanation
of the user flow details by the deploy mode (client vs cluster).
STS is a server and its best way of deployment in K8S cluster is either
done through the helm chart or through the yaml file(although it can be done
through the method you had suggested, but i guess that scenario would be a rare
case and there will be no HA of the STS server if it is triggered from outside).
>> In the scenario of a cluster-mode submission, what is the command-line
behavior? Does the thrift-server script "block" until the thrift server pod is
shut down?
By default the script returns but can be made to block by setting the
environment variable SPARK_NO_DAEMONIZE. Once this is done, script blocks until
the thrift server pod is shut down
>>If possible, there should be some basic integration testing. Run a thrift
server command against the minishift cluster used by the other testing.
Will add it as a separate PR.
Pls merge this, if you are ok with the responses.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
