I think the callback url is not reachable by all the nodes on your hadoop cluster. Since, the callback url fails, there is no way for the workers to notify the oozie server that the action has finished execution. After 10 mins, the oozie server polls the worker node to get the status. Thats why you are seeing 10 mins pauses. Set the following variable in oozie-env.sh to the ipaddres of the server instead of the hostname and restart the oozie server export OOZIE_HTTP_HOSTNAME=<server ip>
On Fri, Jun 27, 2014 at 10:19 AM, Jonathan Hodges <[email protected]> wrote: > Hi, > > We have noticed some pauses between execution of action steps in our Oozie > workflows. A job may take a couple minutes yet the next step isn't > triggered for sometimes ten minutes. Is there a way to control this pause? > We have a set of jobs that when executed at a command-line runs for about > 45 mins. The same set of jobs in Oozie is over 3 hours. > > We tried setting the following properties, but we aren't using coordinators > so it didn't help. > > oozie.service.CoordMaterializeTriggerService.lookup.interval=30 > oozie.service.CoordMaterializeTriggerService.materialization.window=7200 > > Is there maybe similar properties at the workflow level we can set? > > Thanks so much, > > Jonathan >
