[ https://issues.apache.org/jira/browse/TWILL-61?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16393689#comment-16393689 ]
ASF GitHub Bot commented on TWILL-61: ------------------------------------- GitHub user chtyim opened a pull request: https://github.com/apache/twill/pull/67 (TWILL-61) Fix to allow higher attempts to relaunch the app after the first attempt failed - Delete the Kafka root zk node for the application if already exist - Delete the AM instance zk node if already exist - For runnables parent zk node, it is not an error if it already exist - Enhance KafkaClient publisher / consumer to deal with Kafka cluster changes - When AM killed and restarted, the embedded Kafka will be running in different host and port You can merge this pull request into a Git repository by running: $ git pull https://github.com/chtyim/twill feature/TWILL-61 Alternatively you can review and apply these changes as the patch at: https://github.com/apache/twill/pull/67.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #67 ---- commit 9decca071a9e067b30be2150a6097463c939b6af Author: Terence Yim <chtyim@...> Date: 2018-03-09T20:21:26Z (TWILL-61) Fix to allow higher attempts to relaunch the app after the first attempt failed - Delete the Kafka root zk node for the application if already exist - Delete the AM instance zk node if already exist - For runnables parent zk node, it is not an error if it already exist - Enhance KafkaClient publisher / consumer to deal with Kafka cluster changes - When AM killed and restarted, the embedded Kafka will be running in different host and port ---- > Second launch attempt of AM always failed > ----------------------------------------- > > Key: TWILL-61 > URL: https://issues.apache.org/jira/browse/TWILL-61 > Project: Apache Twill > Issue Type: Bug > Components: yarn > Reporter: Terence Yim > Assignee: Terence Yim > Priority: Major > Fix For: 0.5.0-incubating > > > YARN would make multiple attempts to launch an application. Currently second > or above attempts would always fail due to creation of /runId/state node in > ZK fail (node exists) because runId is generated on client side and doesn't > change between attempts. -- This message was sent by Atlassian JIRA (v7.6.3#76005)