Le 4 janv. 2015 00:01, "David Morel" <[email protected]> a écrit :
>
> At the end of your A workflow, use an FS action to set a flag in a
date/time named directory somewhere in HDFS, and use that in an input
synchronous dataset in your B coordinator. It will wait on your A workflow
completion. Or use zookeeper.
>
> David

I forgot to mention : you can have B coordinators have a set timeout, so if
for a given date/time A didn't run then don't wait forever. This is all
based on the assumption your A workflow runs (or doesn't run) at regular
intervals, like every 10 minutes. Or you could use a single file (no
date/time notion) used as a startup flag for workflow B. In that case, as
soon as B starts it deletes the file, and runs. Next instances of B would
just wait (provided concurrency is set to 1) or die based on the timeout
you set. As I write this I find it a bit clunky, and I think a ZK based
solution would be cleaner :-)

>
> Le 3 janv. 2015 15:38, "Laurent H" <[email protected]> a écrit :
>
>> Hi Anil,
>>
>> I don't have any experience with the following solution but maybe, can
help
>> you :
>>
https://oozie.apache.org/docs/3.2.0-incubating/WorkflowFunctionalSpec.html#a3.2.6_Sub-workflow_Action
>>
>> Best Regards,
>>
>> --
>> Laurent HATIER - Consultant Big Data & Business Intelligence chez
CapGemini
>> fr.linkedin.com/pub/laurent-hatier/25/36b/a86/
>> <http://fr.linkedin.com/pub/laurent-h/25/36b/a86/>
>>
>> 2014-12-30 1:37 GMT+01:00 anil gupta <[email protected]>:
>>
>> > Hi,
>> >
>> > I have two oozie workflows A and B. Output of A is input to B. A runs
many
>> > times a day. I would like to run B everytime there is a successful run
of
>> > A.
>> > I can do this very easily if i combine A and B into one workflow.
>> > Unfortunately, i cant do that.
>> > I am still new to Oozie, It seems like i can run B after every
successful
>> > run of A using Oozie Coordinator. Can anyone point me to an example or
>> > provide me some hints of how to achieve this in oozie?
>> >
>> > --
>> > Thanks & Regards,
>> > Anil Gupta
>> >

Reply via email to