ConfX created HDFS-17102:
----------------------------
Summary: Timeout encountered when running
TestDataNodeOutlierDetectionViaMetrics
Key: HDFS-17102
URL: https://issues.apache.org/jira/browse/HDFS-17102
Project: Hadoop HDFS
Issue Type: Bug
Reporter: ConfX
Attachments: reproduce.sh
h2. What happened:
Got a timeout when running {{TestDataNodeOutlierDetectionViaMetrics}} and
setting min outlier to 0 or negative.
h2. Where's the bug:
In {{TestDataNodeOutlierDetectionViaMetrics.injectFastNodesSamples}} the test
injects several packets into the nodes:
{noformat}
for (int i = 0;
i < 2 * peerMetrics.getMinOutlierDetectionSamples();
++i) {
peerMetrics.addSendPacketDownstream(
nodeName, random.nextInt(FAST_NODE_MAX_LATENCY_MS));
}{noformat}
A similar logic appears in the {{{}injectSlowNodesSamples{}}}. A problem with
this code is that if
{{dfs.datanode.peer.metrics.min.outlier.detection.samples}} is set to negative
or 0, no packet would be injected and the {{waitFor}} later:
{noformat}
GenericTestUtils.waitFor(new Supplier<Boolean>() {
@Override
public Boolean get() {
return peerMetrics.getOutliers().size() > 0;
}
}, 500, 100_000);{noformat}
would keeping waiting until timeout.
h2. How to reproduce:
(1) Set {{dfs.datanode.peer.metrics.min.outlier.detection.samples }} to {{0}}
(2) Run test:
{{org.apache.hadoop.hdfs.server.datanode.metrics.TestDataNodeOutlierDetectionViaMetrics#testOutlierIsDetected}}
h2. Stacktrace:
{noformat}
java.util.concurrent.TimeoutException:
Timed out waiting for condition.
Thread diagnostics:
Timestamp: 2023-07-04 04:08:54,535
"Reference Handler" daemon prio=10 tid=2 runnable
java.lang.Thread.State: RUNNABLE
at
[email protected]/java.lang.ref.Reference.waitForReferencePendingList(Native
Method)
at
[email protected]/java.lang.ref.Reference.processPendingReferences(Reference.java:241)
at
[email protected]/java.lang.ref.Reference$ReferenceHandler.run(Reference.java:213)
"surefire-forkedjvm-command-thread" daemon prio=5 tid=23 runnable
java.lang.Thread.State: RUNNABLE
...
{noformat}
For an easy reproduction, run the reproduce.sh in the attachment.
We are happy to provide a patch if this issue is confirmed.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
