Jarek Jarcec Cecho created SQOOP-1816:
-----------------------------------------
Summary: Sqoop2: Add logging to the test DerbyProvider
implementation
Key: SQOOP-1816
URL: https://issues.apache.org/jira/browse/SQOOP-1816
Project: Sqoop
Issue Type: Bug
Affects Versions: 1.99.4
Reporter: Jarek Jarcec Cecho
Assignee: Jarek Jarcec Cecho
Fix For: 1.99.5
I was recently investigating why integration tests are in infinite loop on my
machine. I've seen only a lot of repeating messages from HDFS mini cluster
talking about heartbeats and nothing else was happening. I originally though
that something is going on in the Hadoop minicluster.
Looking closer at {{jstack}} dump I've noticed that the main thread is stuck in
{{[DriverManager.getConnection()|https://github.com/apache/sqoop/blob/sqoop2/common-test/src/main/java/org/apache/sqoop/common/test/db/DatabaseProvider.java#L131]}},
waiting on opening connection to testing Derby database. The Derby should have
been [already
started|https://github.com/apache/sqoop/blob/sqoop2/common-test/src/main/java/org/apache/sqoop/common/test/db/DerbyProvider.java#L39]
at that point in execution, so the wait shouldn't be there. Sadly there
weren't any logs from Derby, so It was hard to figure out what is happening.
I've eventually triaged the issue to port 1527 being sort of used and hence
Derby wasn't able to start properly on that port. This seems as expected
behavior that is covered by DERBY-1465 - unsuccessful Derby start won't thrown
an exception as anticipated, one needs to explicitly call method {{ping()}} to
verify that the server is indeed up and running.
Triaging this issue would be much simpler if we would be logging Derby output
as I would get the exception that derby server can't start in the logs. Hence I
would like to suggest to add that. It would be also helpful to explicitly call
the {{ping()}} method to die fast rather then end up in infinite loop.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
