big-data/hadoop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

YARN-7529. TestYarnNativeServices#testRecoverComponentsAfterRMRestart() fails intermittently. Contributed by Chandni Singh

Browse Source
This commit is contained in:
Billie Rinaldi 2017-11-20 07:37:04 -08:00
parent 6903cf096e
commit 6f9d7a146d
1 changed files with 18 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-services/hadoop-yarn-services-core/src/test/java/org/apache/hadoop/yarn/service/TestYarnNativeServices.java
View File

@ -176,7 +176,8 @@ public void testRecoverComponentsAfterRMRestart() throws Exception {
ServiceClient client = createClient(); ServiceClient client = createClient();
Service exampleApp = createExampleApplication(); Service exampleApp = createExampleApplication();
client.actionCreate(exampleApp); client.actionCreate(exampleApp);
waitForAllCompToBeReady(client, exampleApp); Multimap<String, String> containersBeforeFailure =
waitForAllCompToBeReady(client, exampleApp);
LOG.info("Restart the resource manager"); LOG.info("Restart the resource manager");
getYarnCluster().restartResourceManager( getYarnCluster().restartResourceManager(
@ -191,9 +192,6 @@ public void testRecoverComponentsAfterRMRestart() throws Exception {
ApplicationAttemptId applicationAttemptId = client.getYarnClient() ApplicationAttemptId applicationAttemptId = client.getYarnClient()
.getApplicationReport(exampleAppId).getCurrentApplicationAttemptId(); .getApplicationReport(exampleAppId).getCurrentApplicationAttemptId();
Multimap<String, String> containersBeforeFailure = getContainersForAllComp(
client, exampleApp);
LOG.info("Fail the application attempt {}", applicationAttemptId); LOG.info("Fail the application attempt {}", applicationAttemptId);
client.getYarnClient().failApplicationAttempt(applicationAttemptId); client.getYarnClient().failApplicationAttempt(applicationAttemptId);
//wait until attempt 2 is running //wait until attempt 2 is running
@ -208,7 +206,7 @@ public void testRecoverComponentsAfterRMRestart() throws Exception {
} }
}, 2000, 200000); }, 2000, 200000);
Multimap<String, String> containersAfterFailure = getContainersForAllComp( Multimap<String, String> containersAfterFailure = waitForAllCompToBeReady(
client, exampleApp); client, exampleApp);
Assert.assertEquals("component container affected by restart", Assert.assertEquals("component container affected by restart",
containersBeforeFailure, containersAfterFailure); containersBeforeFailure, containersAfterFailure);
@ -318,14 +316,26 @@ private void waitForOneCompToBeReady(ServiceClient client,
}, 2000, 200000); }, 2000, 200000);
} }
// wait until all the containers for all components become ready state /**
private void waitForAllCompToBeReady(ServiceClient client, * Wait until all the containers for all components become ready state.
*
* @param client
* @param exampleApp
* @return all ready containers of a service.
* @throws TimeoutException
* @throws InterruptedException
*/
private Multimap<String, String> waitForAllCompToBeReady(ServiceClient client,
Service exampleApp) throws TimeoutException, InterruptedException { Service exampleApp) throws TimeoutException, InterruptedException {
int expectedTotalContainers = countTotalContainers(exampleApp); int expectedTotalContainers = countTotalContainers(exampleApp);
Multimap<String, String> allContainers = HashMultimap.create();
GenericTestUtils.waitFor(() -> { GenericTestUtils.waitFor(() -> {
try { try {
Service retrievedApp = client.getStatus(exampleApp.getName()); Service retrievedApp = client.getStatus(exampleApp.getName());
int totalReadyContainers = 0; int totalReadyContainers = 0;
allContainers.clear();
LOG.info("Num Components " + retrievedApp.getComponents().size()); LOG.info("Num Components " + retrievedApp.getComponents().size());
for (Component component : retrievedApp.getComponents()) { for (Component component : retrievedApp.getComponents()) {
LOG.info("looking for " + component.getName()); LOG.info("looking for " + component.getName());
@ -339,6 +349,7 @@ private void waitForAllCompToBeReady(ServiceClient client,
+ component.getName()); + component.getName());
if (container.getState() == ContainerState.READY) { if (container.getState() == ContainerState.READY) {
totalReadyContainers++; totalReadyContainers++;
allContainers.put(component.getName(), container.getId());
LOG.info("Found 1 ready container " + container.getId()); LOG.info("Found 1 ready container " + container.getId());
} }
} }
@ -358,23 +369,6 @@ private void waitForAllCompToBeReady(ServiceClient client,
return false; return false;
} }
}, 2000, 200000); }, 2000, 200000);
}
/**
* Get all containers of a service.
*/
private Multimap<String, String> getContainersForAllComp(ServiceClient client,
Service example) throws IOException, YarnException {
Multimap<String, String> allContainers = HashMultimap.create();
Service retrievedApp = client.getStatus(example.getName());
retrievedApp.getComponents().forEach(component -> {
if (component.getContainers() != null) {
component.getContainers().forEach(container -> {
allContainers.put(component.getName(), container.getId());
});
}
});
return allContainers; return allContainers;
} }