diff options
| author | Sandy Ryza <sandy@cloudera.com> | 2014-02-28 09:40:47 -0600 |
|---|---|---|
| committer | Thomas Graves <tgraves@apache.org> | 2014-02-28 09:40:47 -0600 |
| commit | 5f419bf9f433e8f057237f1d5bfed9f5f4e9427c (patch) | |
| tree | 796cda45fed79d96e6364a42880f4abd1f08e430 /yarn/stable | |
| parent | edf8a56ab7eaee1f7c3b4579eb10464984d31d7a (diff) | |
| download | spark-5f419bf9f433e8f057237f1d5bfed9f5f4e9427c.tar.gz spark-5f419bf9f433e8f057237f1d5bfed9f5f4e9427c.tar.bz2 spark-5f419bf9f433e8f057237f1d5bfed9f5f4e9427c.zip | |
SPARK-1032. If Yarn app fails before registering, app master stays aroun...
...d long after
This reopens https://github.com/apache/incubator-spark/pull/648 against the new repo.
Author: Sandy Ryza <sandy@cloudera.com>
Closes #28 from sryza/sandy-spark-1032 and squashes the following commits:
5953f50 [Sandy Ryza] SPARK-1032. If Yarn app fails before registering, app master stays around long after
Diffstat (limited to 'yarn/stable')
| -rw-r--r-- | yarn/stable/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala | 22 |
1 files changed, 16 insertions, 6 deletions
diff --git a/yarn/stable/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala b/yarn/stable/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala index 4b777d5fa7..0f58c49c69 100644 --- a/yarn/stable/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala +++ b/yarn/stable/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala @@ -67,6 +67,8 @@ class ApplicationMaster(args: ApplicationMasterArguments, conf: Configuration, private val maxNumWorkerFailures = sparkConf.getInt("spark.yarn.max.worker.failures", math.max(args.numWorkers * 2, 3)) + private var registered = false + def run() { // Setup the directories so things go to YARN approved directories rather // than user specified and /tmp. @@ -99,7 +101,12 @@ class ApplicationMaster(args: ApplicationMasterArguments, conf: Configuration, waitForSparkContextInitialized() // Do this after Spark master is up and SparkContext is created so that we can register UI Url. - val appMasterResponse: RegisterApplicationMasterResponse = registerApplicationMaster() + synchronized { + if (!isFinished) { + registerApplicationMaster() + registered = true + } + } // Allocate all containers allocateWorkers() @@ -180,7 +187,8 @@ class ApplicationMaster(args: ApplicationMasterArguments, conf: Configuration, var numTries = 0 val waitTime = 10000L val maxNumTries = sparkConf.getInt("spark.yarn.applicationMaster.waitTries", 10) - while (ApplicationMaster.sparkContextRef.get() == null && numTries < maxNumTries) { + while (ApplicationMaster.sparkContextRef.get() == null && numTries < maxNumTries + && !isFinished) { logInfo("Waiting for Spark context initialization ... " + numTries) numTries = numTries + 1 ApplicationMaster.sparkContextRef.wait(waitTime) @@ -313,11 +321,13 @@ class ApplicationMaster(args: ApplicationMasterArguments, conf: Configuration, return } isFinished = true - } - logInfo("finishApplicationMaster with " + status) - // Set tracking URL to empty since we don't have a history server. - amClient.unregisterApplicationMaster(status, "" /* appMessage */ , "" /* appTrackingUrl */) + logInfo("finishApplicationMaster with " + status) + if (registered) { + // Set tracking URL to empty since we don't have a history server. + amClient.unregisterApplicationMaster(status, "" /* appMessage */ , "" /* appTrackingUrl */) + } + } } /** |