1255457 - Presto connection closed to Hive

Status: RESOLVED FIXED

(Cloud Services Graveyard :: Metrics: Pipeline, defect, P1)

Description

[Roberto Agostino Vitillo (:rvitillo)]

See [1] for the error and a possible solution. Not sure we started experiencing this issue only now.

[1] https://forums.aws.amazon.com/thread.jspa?threadID=220302

Comment 1

[Wesley Dawson [:whd]]

We're testing the fix outlined in comment #1. For our flavor of Presto/Hive the classpaths are such that:

ln -s /usr/lib/presto/plugin/mysql/mysql-connector-java-5.1.35.jar /usr/lib/hadoop/lib/mysql-connector-java.jar

DTRT. We can make this fix permanent using techniques in https://docs.aws.amazon.com/ElasticMapReduce/latest/DeveloperGuide/emr-dev-create-metastore-outside.html, but at that point we should probably go whole hog and make the related change of storing this in a proper RDS instance.

A workaround we discovered is to run hive periodically.

Comment 2

[Mark Finkle (:mfinkle) (use needinfo?)]

I was running a non-trivial amount of queries through Redash over the weekend (3/12 and 3/13) and did not hit the exception. Seems like the workaround is working.

Comment 3

[Thomas Huelbert]

back to a P2 - :whd is on ops related tasks for the next while. will prioritize when when he becomes available again.

Comment 4

[Wesley Dawson [:whd]]

This seems to have been fixed by upgrading to Hive 1.2.