Marvin Rösch created ZEPPELIN-5562:
--------------------------------------
Summary: Recovered remote interpreter instance does not update
correct resource pool
Key: ZEPPELIN-5562
URL: https://issues.apache.org/jira/browse/ZEPPELIN-5562
Project: Zeppelin
Issue Type: Bug
Affects Versions: 0.10.0
Environment: Zeppelin on Kubernetes with k8s interpreter launcher. We
isolate our Python and R interpreters per user, while interpreters such as our
JDBC one to query our database are shared for everyone.
Reporter: Marvin Rösch
We run Zeppelin on Kubernetes using the k8s interpreter launcher. We also have
interpreter recovery activated, so Pods for interpreters don't get deleted when
their parent Zeppelin Pod gets stopped, but instead are kept around to enable
quick recovery after the main server was restarted.
After a reconnect from the server to an interpreter, the {{saveAs}} option does
not appear to work correctly anymore. Saving the result of an SQL query from a
recovered JDBC interpreter in a variable does not make it available in e.g. a
Python interpreter - regardless of that interpreter being recovered or not.
I investigated the way interpreter recovery works in the code and from what I
saw it _should_ be properly updating the {{DistributedResourcePool}} after a
reconnect, but this does not seem to occur or the new connection isn't used for
{{saveAs}}.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
