JIRA ticket issued:
https://issues.apache.org/jira/browse/OOZIE-1222

Thanks,
Shanzhong


On Tue, Feb 12, 2013 at 2:45 PM, Mona Chitnis <chit...@yahoo-inc.com> wrote:

> Seems like a fair enhancement. Could you please file a JIRA stating these
> details?
>
> Thanks,
>
> Mona
>
> On 2/11/13 6:03 PM, "Shangzhong zhu" <shanzh...@gmail.com> wrote:
>
> >Thanks, Mona.
> >
> >I could rerun the workflow job, but the corresponding coordinator action
> >status won't get changed, right? which means the coordinator action will
> >still show failed, even if the associated workflow get rerun successfully.
> >
> >Can we enhance the coordinator rerun to be consistent with the workflow
> >rerun?
> >* Keep the same workflow ID.
> >* Support rerun from beginning or rerun from the failed WF action.
> >* Rerun count reflects the number of tries.
> >
> >Thanks,
> >Shanzhong
> >
> >On Mon, Feb 11, 2013 at 1:11 PM, Mona Chitnis <chit...@yahoo-inc.com>
> >wrote:
> >
> >> Hi folks,
> >>
> >> Looking into the Coordinator Rerun logic, it looks like rerunning a
> >> coordinator action resets its external id (which maps to workflow job)
> >>and
> >> external status. This means it will run a fresh workflow job which
> >> explains why the client method getRerun() was returning '0'.
> >>
> >> For using the max-rerun limit, you can use OozieClient.rerun() method
> >> itself and supply to it the workflow job-id obtained from coordinator
> >> action's externalId.
> >>
> >> Thanks,
> >>
> >> Mona
> >>
> >> On 2/10/13 11:12 PM, "Shangzhong zhu" <shanzh...@gmail.com> wrote:
> >>
> >> >Thanks, Alejandro.
> >> >
> >> >For the WF rerun count you mentioned, is it
> >> >org.apache.oozie.clien.WorkflowJob.getReRun()?
> >> >
> >> >However, it seems always return 0 no matter how many rerun I made by
> >>using
> >> >Coordinator Rerun.
> >> >
> >> >Basically, I am using coordinator rerun: OozieClient.coordReRun() to
> >>rerun
> >> >failed coordinator actions/workflows. But I want to control the number
> >>of
> >> >reruns, say maximum 3 reruns.
> >> >
> >> >Thanks,
> >> >Shanzhong
> >> >
> >> >
> >> >On Mon, Feb 4, 2013 at 4:17 PM, Alejandro Abdelnur
> >> ><t...@cloudera.com>wrote:
> >> >
> >> >> On Sat, Feb 2, 2013 at 11:05 PM, Shangzhong zhu <shanzh...@gmail.com
> >
> >> >> wrote:
> >> >>
> >> >> > Hi All,
> >> >> >
> >> >> > We are developing a wrapper on top of oozie to automate
> >>failed/killed
> >> >> > coordinator action rerun.
> >> >> >
> >> >> > To rerun a coordinator action, seems I have two options.
> >> >> >
> >> >> > 1. Using coordinator action rerun:
> >> >> >      oozie job -rerun <coord_Job_id> <-date XXXX>
> >> >> >
> >> >> > 2. Since the failed action is a workflow job, I can also rerun that
> >> >> > workflow job by setting oozie.wf.rerun.failnodes to rerun from the
> >> >>failed
> >> >> > action.
> >> >> >
> >> >> > Questions:
> >> >> > 0. which option is preferred?
> >> >> >
> >> >> > 1. For option 1, can I choose to rerun from the failed action like
> >>the
> >> >> > oozie.wf.rerun.failnodes option in workflow rerun?
> >> >> >
> >> >> > If I recall correctly you cannot do this.
> >> >>
> >> >>
> >> >> > 2. For option 1, seems I cannot change the job configurations. But
> >>for
> >> >> > option 2, I have more flexibility in changing the configurations,
> >>say
> >> >>I
> >> >> can
> >> >> > change the job name so that I know how many rerun has been made for
> >> >>that
> >> >> > workflow.
> >> >> >
> >> >> > no need for this, there is a WF rerun count.
> >> >>
> >> >>
> >> >> > 3. If I chose option 2, does it mean that the rerun workflow job is
> >> >>not
> >> >> > part of the coordinator actions any more? In another word, if I
> >>killed
> >> >> that
> >> >> > coordinator job, that rerun workflow job will be still running?
> >> >>
> >> >>
> >> >> It should get killed as well as the WF job ID is still the same as.
> >> >>
> >> >> Wit Option #2 though I'm not sure what will happen with the status of
> >> >>the
> >> >> corresponding COORD action.
> >> >>
> >> >>
> >> >>
> >> >> >
> >> >> >
> >> >> Thanks
> >> >> >
> >> >>
> >> >>
> >> >>
> >> >> --
> >> >> Alejandro
> >> >>
> >>
> >>
>
>

Reply via email to