Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

YARN-8910. Fixed misleading log statement when container max retries is infinite. 

Contributed by Chandni Singh
This commit is contained in:
Eric Yang 2018-10-19 13:49:04 -04:00
parent 88cce32551
commit 47ad98b2e1
3 changed files with 37 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/container/ContainerImpl.java
View File

@ -1653,10 +1653,17 @@ public ContainerState transition(final ContainerImpl container,
private void doRelaunch(final ContainerImpl container,
int remainingRetryAttempts, final int retryInterval) {
LOG.info("Relaunching Container " + container.getContainerId()
+ ". Remaining retry attempts(after relaunch) : "
+ remainingRetryAttempts + ". Interval between retries is "
+ retryInterval + "ms");
if (remainingRetryAttempts == ContainerRetryContext.RETRY_FOREVER) {
LOG.info("Relaunching Container {}. " +
"retry interval {} ms", container.getContainerId(),
retryInterval);
} else {
LOG.info("Relaunching Container {}. " +
"remaining retry attempts(after relaunch) {}, " +
"retry interval {} ms", container.getContainerId(),
remainingRetryAttempts, retryInterval);
}
container.wasLaunched = false;
container.metrics.endRunningContainer();
if (retryInterval == 0) {

4
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/container/SlidingWindowRetryPolicy.java
View File

@ -153,6 +153,10 @@ ContainerRetryContext getContainerRetryContext() {
}
int getRemainingRetries() {
if (containerRetryContext.getMaxRetries() ==
ContainerRetryContext.RETRY_FOREVER) {
return ContainerRetryContext.RETRY_FOREVER;
}
return remainingRetries;
}

26
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/test/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/container/TestSlidingWindowRetryPolicy.java
View File

@ -43,8 +43,12 @@ public void setup() {
public void testNeverRetry() {
ContainerRetryContext retryContext =
ContainerRetryContext.NEVER_RETRY_CONTEXT;
Assert.assertFalse("never retry", retryPolicy.shouldRetry(
new SlidingWindowRetryPolicy.RetryContext(retryContext), 12));
SlidingWindowRetryPolicy.RetryContext windowContext = new
SlidingWindowRetryPolicy.RetryContext(retryContext);
Assert.assertFalse("never retry", retryPolicy.shouldRetry(windowContext,
12));
Assert.assertEquals("remaining retries", 0,
windowContext.getRemainingRetries());
}
@Test
@ -52,8 +56,13 @@ public void testAlwaysRetry() {
ContainerRetryContext retryContext = ContainerRetryContext.newInstance(
ContainerRetryPolicy.RETRY_ON_ALL_ERRORS, null, -1,
0, 10);
Assert.assertTrue("always retry", retryPolicy.shouldRetry(
new SlidingWindowRetryPolicy.RetryContext(retryContext), 12));
SlidingWindowRetryPolicy.RetryContext windowContext = new
SlidingWindowRetryPolicy.RetryContext(retryContext);
Assert.assertTrue("always retry", retryPolicy.shouldRetry(windowContext,
12));
Assert.assertEquals("remaining retries",
ContainerRetryContext.RETRY_FOREVER,
windowContext.getRemainingRetries());
}
@Test
@ -65,19 +74,28 @@ public void testFailuresValidityInterval() {
Assert.assertTrue("retry 1",
retryPolicy.shouldRetry(windowRetryContext, 12));
retryPolicy.updateRetryContext(windowRetryContext);
Assert.assertEquals("remaining retries", 1,
windowRetryContext.getRemainingRetries());
clock.setTime(20);
Assert.assertTrue("retry 2",
retryPolicy.shouldRetry(windowRetryContext, 12));
retryPolicy.updateRetryContext(windowRetryContext);
Assert.assertEquals("remaining retries", 1,
windowRetryContext.getRemainingRetries());
clock.setTime(40);
Assert.assertTrue("retry 3",
retryPolicy.shouldRetry(windowRetryContext, 12));
retryPolicy.updateRetryContext(windowRetryContext);
Assert.assertEquals("remaining retries", 1,
windowRetryContext.getRemainingRetries());
clock.setTime(45);
Assert.assertFalse("retry failed",
retryPolicy.shouldRetry(windowRetryContext, 12));
retryPolicy.updateRetryContext(windowRetryContext);
Assert.assertEquals("remaining retries", 0,
windowRetryContext.getRemainingRetries());
}
}