From a4041dd87f7b33b28de29ef0a4eebe33c7b0e6ca Mon Sep 17 00:00:00 2001 From: Charles Reiss Date: Thu, 13 Dec 2012 16:11:08 -0800 Subject: Log duplicate slaveLost() calls in ClusterScheduler. --- .../src/main/scala/spark/scheduler/cluster/ClusterScheduler.scala | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/core/src/main/scala/spark/scheduler/cluster/ClusterScheduler.scala b/core/src/main/scala/spark/scheduler/cluster/ClusterScheduler.scala index d160379b14..ab200decb1 100644 --- a/core/src/main/scala/spark/scheduler/cluster/ClusterScheduler.scala +++ b/core/src/main/scala/spark/scheduler/cluster/ClusterScheduler.scala @@ -254,14 +254,20 @@ private[spark] class ClusterScheduler(val sc: SparkContext) synchronized { val host = slaveIdToHost(slaveId) if (hostsAlive.contains(host)) { + logError("Lost an executor on " + host + ": " + reason) slaveIdsWithExecutors -= slaveId hostsAlive -= host activeTaskSetsQueue.foreach(_.hostLost(host)) failedHost = Some(host) + } else { + // We may get multiple slaveLost() calls with different loss reasons. For example, one + // may be triggered by a dropped connection from the slave while another may be a report + // of executor termination from Mesos. We produce log messages for both so we eventually + // report the termination reason. + logError("Lost an executor on " + host + " (already removed): " + reason) } } if (failedHost != None) { - logError("Lost an executor on " + failedHost.get + ": " + reason) listener.hostLost(failedHost.get) backend.reviveOffers() } -- cgit v1.2.3