Repository: hadoop Updated Branches: refs/heads/trunk 9130af3f7 -> 8a0fa0f7e
HDFS-10733. NameNode terminated after full GC thinking QJM is unresponsive. Contributed by Vinitha Gankidi. Project: http://git-wip-us.apache.org/repos/asf/hadoop/repo Commit: http://git-wip-us.apache.org/repos/asf/hadoop/commit/8a0fa0f7 Tree: http://git-wip-us.apache.org/repos/asf/hadoop/tree/8a0fa0f7 Diff: http://git-wip-us.apache.org/repos/asf/hadoop/diff/8a0fa0f7 Branch: refs/heads/trunk Commit: 8a0fa0f7e88c45a98c6f266d6349cb426dd06495 Parents: 9130af3 Author: Vinitha Reddy Gankidi <vigank...@linkedin.com> Authored: Tue Jan 17 17:21:12 2017 -0800 Committer: Konstantin V Shvachko <s...@apache.org> Committed: Wed Jan 18 12:46:32 2017 -0800 ---------------------------------------------------------------------- .../hadoop/hdfs/qjournal/client/QuorumCall.java | 26 +++++++++++++++++++- .../hdfs/qjournal/client/TestQuorumCall.java | 17 +++++++++++++ 2 files changed, 42 insertions(+), 1 deletion(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/hadoop/blob/8a0fa0f7/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/QuorumCall.java ---------------------------------------------------------------------- diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/QuorumCall.java b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/QuorumCall.java index f15e462..dc32318 100644 --- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/QuorumCall.java +++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/QuorumCall.java @@ -20,8 +20,10 @@ package org.apache.hadoop.hdfs.qjournal.client; import java.util.Map; import java.util.Map.Entry; import java.util.concurrent.TimeoutException; +import java.util.concurrent.TimeUnit; import org.apache.hadoop.ipc.RemoteException; +import org.apache.hadoop.util.StopWatch; import org.apache.hadoop.util.Time; import com.google.common.base.Joiner; @@ -58,6 +60,7 @@ class QuorumCall<KEY, RESULT> { * fraction of the configured timeout for any call. */ private static final float WAIT_PROGRESS_WARN_THRESHOLD = 0.7f; + private final StopWatch quorumStopWatch = new StopWatch(); static <KEY, RESULT> QuorumCall<KEY, RESULT> create( Map<KEY, ? extends ListenableFuture<RESULT>> calls) { @@ -83,6 +86,16 @@ class QuorumCall<KEY, RESULT> { private QuorumCall() { // Only instantiated from factory method above } + + private void restartQuorumStopWatch() { + quorumStopWatch.reset().start(); + } + + private boolean shouldIncreaseQuorumTimeout(long offset, int millis) { + long elapsed = quorumStopWatch.now(TimeUnit.MILLISECONDS); + return elapsed + offset > (millis * WAIT_PROGRESS_INFO_THRESHOLD); + } + /** * Wait for the quorum to achieve a certain number of responses. @@ -110,6 +123,7 @@ class QuorumCall<KEY, RESULT> { long nextLogTime = st + (long)(millis * WAIT_PROGRESS_INFO_THRESHOLD); long et = st + millis; while (true) { + restartQuorumStopWatch(); checkAssertionErrors(); if (minResponses > 0 && countResponses() >= minResponses) return; if (minSuccesses > 0 && countSuccesses() >= minSuccesses) return; @@ -139,11 +153,21 @@ class QuorumCall<KEY, RESULT> { } long rem = et - now; if (rem <= 0) { - throw new TimeoutException(); + // Increase timeout if a full GC occurred after restarting stopWatch + if (shouldIncreaseQuorumTimeout(0, millis)) { + et = et + millis; + } else { + throw new TimeoutException(); + } } + restartQuorumStopWatch(); rem = Math.min(rem, nextLogTime - now); rem = Math.max(rem, 1); wait(rem); + // Increase timeout if a full GC occurred after restarting stopWatch + if (shouldIncreaseQuorumTimeout(-rem, millis)) { + et = et + millis; + } } } http://git-wip-us.apache.org/repos/asf/hadoop/blob/8a0fa0f7/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/client/TestQuorumCall.java ---------------------------------------------------------------------- diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/client/TestQuorumCall.java b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/client/TestQuorumCall.java index 2295384..506497e 100644 --- a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/client/TestQuorumCall.java +++ b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/client/TestQuorumCall.java @@ -66,4 +66,21 @@ public class TestQuorumCall { // expected } } + @Test(timeout=10000) + public void testQuorumFailsWithoutResponse() throws Exception { + Map<String, SettableFuture<String>> futures = ImmutableMap.of( + "f1", SettableFuture.<String>create()); + + QuorumCall<String, String> q = QuorumCall.create(futures); + assertEquals("The number of quorum calls for which a response has been" + + " received should be 0", 0, q.countResponses()); + + try { + q.waitFor(0, 1, 100, 10, "test"); + fail("Didn't time out waiting for more responses than came back"); + } catch (TimeoutException te) { + // expected + } + } + } --------------------------------------------------------------------- To unsubscribe, e-mail: common-commits-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-commits-h...@hadoop.apache.org