[
https://issues.apache.org/jira/browse/GOBBLIN-744?focusedWorklogId=230595&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-230595
]
ASF GitHub Bot logged work on GOBBLIN-744:
------------------------------------------
Author: ASF GitHub Bot
Created on: 22/Apr/19 05:43
Start Date: 22/Apr/19 05:43
Worklog Time Spent: 10m
Work Description: sv2000 commented on pull request #2609: GOBBLIN-744:
Support cancellation of a Helix workflow via a DELETE Spec.
URL: https://github.com/apache/incubator-gobblin/pull/2609#discussion_r277210433
##########
File path:
gobblin-cluster/src/test/java/org/apache/gobblin/cluster/ClusterIntegrationTest.java
##########
@@ -82,6 +97,72 @@ public void testJobShouldComplete()
suite.waitForAndVerifyOutputFiles();
}
+ /**
+ * An integration test for cancelling a Helix workflow via a JobSpec. This
test case starts a Helix cluster with
+ * a {@link FsScheduledJobConfigurationManager}. The test case does the
following:
+ * <ul>
+ * <li> add a {@link org.apache.gobblin.runtime.api.JobSpec} that uses a
{@link org.apache.gobblin.cluster.SleepingCustomTaskSource})
+ * to {@link IntegrationJobCancelViaSpecSuite#FS_SPEC_CONSUMER_DIR}.
which is picked by the JobConfigurationManager. </li>
+ * <li> the JobConfigurationManager sends a notification to the
GobblinHelixJobScheduler which schedules the job for execution. The JobSpec is
+ * also added to the JobCatalog for persistence. Helix starts a Workflow
for this JobSpec. </li>
+ * <li> We then add a {@link org.apache.gobblin.runtime.api.JobSpec} with
DELETE Verb to {@link IntegrationJobCancelViaSpecSuite#FS_SPEC_CONSUMER_DIR}.
+ * This signals GobblinHelixJobScheduler (and, Helix) to delete the
running job (i.e., Helix Workflow) started in the previous step. </li>
+ * <li> Finally, we inspect the state of the zNode corresponding to the
Workflow resource in Zookeeper to ensure that its {@link
org.apache.helix.task.TargetState}
+ * is STOP. </li>
+ * </ul>
+ */
+ @Test (dependsOnMethods = { "testJobShouldGetCancelled" })
+ public void testJobCancellationViaSpec() throws Exception {
+ this.suite = new IntegrationJobCancelViaSpecSuite();
+ HelixManager helixManager = getHelixManager();
+
+ IntegrationJobCancelViaSpecSuite cancelViaSpecSuite =
(IntegrationJobCancelViaSpecSuite) this.suite;
+
+ //Add a new JobSpec to the path monitored by the SpecConsumer
+ cancelViaSpecSuite.addJobSpec(IntegrationJobCancelViaSpecSuite.JOB_ID,
SpecExecutor.Verb.ADD.name());
+
+ //Start the cluster
+ cancelViaSpecSuite.startCluster();
+
+ helixManager.connect();
+
+ while (TaskDriver.getWorkflowContext(helixManager,
IntegrationJobCancelViaSpecSuite.JOB_ID) == null) {
+ log.warn("Waiting for the job to start...");
+ Thread.sleep(1000L);
+ }
+
+ // Give the job some time to reach writer, where it sleeps
+ Thread.sleep(2000L);
Review comment:
Modified SleepingTask to write an empty file declaring that it is RUNNING.
However, the test case will wait only a bounded amount of time for the
existence of this file, to avoid the pathological condition where the file
cannot be created and the test case keeps polling for the existence of this
file forever.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 230595)
Time Spent: 2h 40m (was: 2.5h)
> Support cancellation of a Helix workflow via a DELETE Spec
> ----------------------------------------------------------
>
> Key: GOBBLIN-744
> URL: https://issues.apache.org/jira/browse/GOBBLIN-744
> Project: Apache Gobblin
> Issue Type: Improvement
> Components: gobblin-cluster
> Affects Versions: 0.15.0
> Reporter: Sudarshan Vasudevan
> Assignee: Hung Tran
> Priority: Major
> Fix For: 0.15.0
>
> Time Spent: 2h 40m
> Remaining Estimate: 0h
>
> This task supports the ability to interrupt and cancel a running job on a
> Gobblin Helix cluster via a DELETE Spec submitted to the
> JobConfigurationManager. The DELETE Spec should have
> "gobblin.cluster.shouldCancelRunningJobOnDelete" set to true for cancelling a
> running job. The default behavior is to simply delete the corresponding
> JobSpec from the JobCatalog.
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
