Oozie coordinator action rerun from fail nodes

Question

I am trying to rerun an oozie co-ordinator action using below command.

oozie job -rerun <co-ordinator id> -action <action id>
-Doozie.wf.rerun.fail.nodes=true

But it is executing the action from the beginning instead of executing it from the point of failed node. Coordinator is neglecting 'oozie.wf.rerun.fail.nodes=true' option. Do I miss to provide some options?

Juetta

Try setting -Doozie.wf.rerun.failnodes=true and see if is solves. – Queenstown

Tried it. No use :( – Juetta

Sorry that is a wrong option. – Queenstown

Answer 1

For Oozie Coordinator rerun,

Use -failed flag which re-runs the failed workflow actions of the coordinator actions that are passed in -action.

oozie job -rerun <co-ordinator id> -failed -action <action id>

The property -Doozie.wf.rerun.failnodes=true is used for rerunning workflows.

Queenstown

But how do we control the coordinator action to run from point of failure instead of running from beginning. – Juetta

You have to pass the failed action ids manually. Coordinator reruns will not look out for failed coordinator actions in itself but can be made to rerun only the failed nodes within the workflows of coordinator options. Refer here for more details. – Queenstown

If in my coordinator if I have action with 10 hive scripts as individual sub-workflows inside it, and if the action fails after running 9 scripts. There is no option that I can continue executing from 9th script? Everytime there is a failure, do I need to start from script 1 sub-workflow again? – Juetta

No, You need not. This scenario is a single action coordinator job with 10 workflows inside. Pass this action id with -failed flag set, the job will run only the failed workflow action, i.e., 9th hive script for the specified action id. – Queenstown