Correct, although some things changed I have not studied up on since Giraph on YARN moved to Hadoop 2.2 compatibility. Eventuallly the smart play will be to move the master to the "application master" process but to maintain easy compatibility with MRv1 Giraph I didn't implement it this way when I put the original path together.
Good luck, hope that helps, Eli On Sun, Mar 2, 2014 at 11:02 PM, Jyotirmoy Sundi <[email protected]> wrote: > Hi Eli, > Please ignore my previous response. > So for Giraph on YARN, are you saying for a large graph when running on > MR1 (runs with one master + multiple worker) but for yarn it would be one > Application Master, one Master, and multiple Workers ? > > Thanks > Sundi > > > > > On Sun, Mar 2, 2014 at 10:14 PM, Jyotirmoy Sundi <[email protected]>wrote: > >> Hmmm, >> I am running in cloudera manager. The number of application master,master >> and worker seems as per the config stats. >> >> I get the following response in master mapper while running on a small >> graph >> MASTER_ONLY checkWorkers: Only found 2 responses of 3 needed to start >> superstep -1 >> >> When I go to the mapper running master, I get the following log: >> >> INFO org.apache.giraph.master.BspServiceMaster: >> logMissingWorkersOnSuperstep: No response from partition 3 (could be master) >> >> Any idea what configuration issue it might be ? >> >> >> Thanks >> >> Sundi >> >> >> >> >> On Sun, Mar 2, 2014 at 4:56 PM, Eli Reisman <[email protected]>wrote: >> >>> This looks like YARN cluster is misconfigured. Alternately, you need to >>> configure it to allow a few more worker tasks. Giraph on YARN at minimum >>> needs one Application Master, one Master, and one Worker (so 3 YARN >>> containers) I have a feeling this could be the issue. >>> >>> >>> On Sat, Mar 1, 2014 at 9:18 PM, Jyotirmoy Sundi <[email protected]>wrote: >>> >>>> Hi Folks, >>>> >>>> The job was working properly in MR1 without any issue. I am trying to >>>> run a simple CC sample Giraph job on YARN. . I have attached the >>>> stacktrace and a few errors. Any pointers will be really helpful for the >>>> below errors. >>>> >>>> *1. BspServiceMaster (YARN profile) is FAILING this task, throwing >>>> exception to end job run.* >>>> >>>> *2. java.lang.IllegalStateException: Not enough healthy workers to create >>>> input splits* >>>> >>>> >>>> >>>> >>>> *StackTrace:* >>>> >>>> 2014-03-02 04:53:24,646 INFO org.apache.giraph.master.BspServiceMaster: >>>> logMissingWorkersOnSuperstep: No response from partition 2 (could be >>>> master) >>>> 2014-03-02 04:53:24,646 ERROR org.apache.giraph.master.BspServiceMaster: >>>> checkWorkers: Did not receive enough processes in time (only 1 of 2 >>>> required) after waiting 600000msecs). This occurs if you do not have >>>> enough map tasks available simultaneously on your Hadoop instance to >>>> fulfill the number of requested workers. >>>> 2014-03-02 04:53:24,649 INFO org.apache.giraph.master.BspServiceMaster: >>>> setJobState: >>>> {"_stateKey":"FAILED","_applicationAttemptKey":-1,"_superstepKey":-1} on >>>> superstep -1 >>>> 2014-03-02 04:53:24,653 FATAL org.apache.giraph.master.BspServiceMaster: >>>> failJob: Killing job job_201402281650_0019 >>>> 2014-03-02 04:53:24,654 FATAL org.apache.giraph.master.BspServiceMaster: >>>> failJob: exception java.lang.IllegalStateException: Not enough healthy >>>> workers to create input splits >>>> 2014-03-02 04:53:24,654 ERROR org.apache.giraph.master.MasterThread: >>>> masterThread: Master algorithm failed with RuntimeException >>>> java.lang.RuntimeException: BspServiceMaster (YARN profile) is FAILING >>>> this task, throwing exception to end job run. >>>> at >>>> org.apache.giraph.master.BspServiceMaster.failJob(BspServiceMaster.java:349) >>>> at >>>> org.apache.giraph.master.BspServiceMaster.setJobStateFailed(BspServiceMaster.java:297) >>>> at >>>> org.apache.giraph.master.BspServiceMaster.createInputSplits(BspServiceMaster.java:616) >>>> at >>>> org.apache.giraph.master.BspServiceMaster.createVertexInputSplits(BspServiceMaster.java:692) >>>> at org.apache.giraph.master.MasterThread.run(MasterThread.java:100) >>>> Caused by: java.lang.IllegalStateException: Not enough healthy workers to >>>> create input splits >>>> ... 4 more >>>> 2014-03-02 04:53:24,656 FATAL org.apache.giraph.graph.GraphMapper: >>>> uncaughtException: OverrideExceptionHandler on thread >>>> org.apache.giraph.master.MasterThread, msg = java.lang.RuntimeException: >>>> BspServiceMaster (YARN profile) is FAILING this task, throwing exception >>>> to end job run., exiting... >>>> java.lang.IllegalStateException: java.lang.RuntimeException: >>>> BspServiceMaster (YARN profile) is FAILING this task, throwing exception >>>> to end job run. >>>> at org.apache.giraph.master.MasterThread.run(MasterThread.java:181) >>>> Caused by: java.lang.RuntimeException: BspServiceMaster (YARN profile) is >>>> FAILING this task, throwing exception to end job run. >>>> at >>>> org.apache.giraph.master.BspServiceMaster.failJob(BspServiceMaster.java:349) >>>> at >>>> org.apache.giraph.master.BspServiceMaster.setJobStateFailed(BspServiceMaster.java:297) >>>> at >>>> org.apache.giraph.master.BspServiceMaster.createInputSplits(BspServiceMaster.java:616) >>>> at >>>> org.apache.giraph.master.BspServiceMaster.createVertexInputSplits(BspServiceMaster.java:692) >>>> at org.apache.giraph.master.MasterThread.run(MasterThread.java:100) >>>> Caused by: java.lang.IllegalStateException: Not enough healthy workers to >>>> create input splits >>>> ... 4 more >>>> >>>> ------------------------------ >>>> >>>> >>> >> >> >> -- >> Best Regards, >> Jyotirmoy Sundi >> >> > > > -- > Best Regards, > Jyotirmoy Sundi > >
