[
https://issues.apache.org/jira/browse/GOBBLIN-2193?focusedWorklogId=955781&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-955781
]
ASF GitHub Bot logged work on GOBBLIN-2193:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 06/Feb/25 07:33
Start Date: 06/Feb/25 07:33
Worklog Time Spent: 10m
Work Description: Blazer-007 commented on code in PR #4096:
URL: https://github.com/apache/gobblin/pull/4096#discussion_r1944213593
##########
gobblin-cluster/src/main/java/org/apache/gobblin/cluster/event/JobFailureEvent.java:
##########
@@ -0,0 +1,20 @@
+package org.apache.gobblin.cluster.event;
+
+import lombok.Getter;
+import org.apache.gobblin.runtime.JobState;
+
+
+/**
+ * The `JobFailureEvent` class represents an event that is triggered when a
job fails.
+ * It contains information about the job state and a summary of the issues
that caused the failure.
+ */
+public class JobFailureEvent {
+ @Getter
+ private final JobState jobState;
+ @Getter
+ private final String issuesSummary;
+ public JobFailureEvent(JobState jobState, String issuesSummary) {
+ this.jobState = jobState;
+ this.issuesSummary = issuesSummary;
+ }
Review Comment:
can we use @AllArgsConstructor in place of this ?
##########
gobblin-temporal/src/main/java/org/apache/gobblin/temporal/joblauncher/GobblinTemporalJobLauncher.java:
##########
@@ -107,13 +111,38 @@ protected Config applyJobLauncherOverrides(Config config)
{
return configOverrides.withFallback(config);
}
+ private String getIssuesSummary() {
+ TextStringBuilder sb = new TextStringBuilder();
+ try {
+ List<Issue> issues = this.getIssueRepository().getAll();
+ sb.appendln("");
+ sb.appendln("vvvvv============= Issues (summary) =============vvvvv");
+
+ for (int i = 0; i < issues.size(); i++) {
+ Issue issue = issues.get(i);
+
+ sb.appendln("%s) %s %s %s | source: %s", i + 1,
issue.getSeverity().toString(), issue.getCode(),
+ issue.getSummary(), issue.getSourceClass());
+ }
+ sb.append("^^^^^=============================================^^^^^");
+ sb.toString();
+ }
+ catch(Exception e) {
+ log.warn("Failed to get issue summary", e);
+ }
+ return sb.toString();
+ }
+
@Override
protected void handleLaunchFinalization() {
// NOTE: This code only makes sense when there is 1 source / workflow
being launched per application for Temporal. This is a stop-gap
// for achieving batch job behavior. Given the current constraints of yarn
applications requiring a static proxy user
// during application creation, it is not possible to have multiple
workflows running in the same application.
// and so it makes sense to just kill the job after this is completed
log.info("Requesting the AM to shutdown after the job {} completed",
this.jobContext.getJobId());
+ JobState jobState = this.jobContext.getJobState();
+ String issuesSummary = this.getIssuesSummary();
+ eventBus.post(new JobFailureEvent(jobState, issuesSummary));
Review Comment:
This event more of a seems like JobIssuesSummaryEvent and not
JobFailureEvent or we can add issues only in case of failure as anyway we have
checked the same while unregistering AM
##########
gobblin-cluster/src/main/java/org/apache/gobblin/cluster/event/JobFailureEvent.java:
##########
@@ -0,0 +1,20 @@
+package org.apache.gobblin.cluster.event;
+
Review Comment:
Please add license statements from other files
##########
gobblin-temporal/src/main/java/org/apache/gobblin/temporal/yarn/YarnService.java:
##########
@@ -198,6 +201,13 @@ class YarnService extends AbstractIdleService {
private final AtomicLong allocationRequestIdGenerator = new
AtomicLong(DEFAULT_ALLOCATION_REQUEST_ID);
private final ConcurrentMap<Long, WorkerProfile>
workerProfileByAllocationRequestId = new ConcurrentHashMap<>();
+ @VisibleForTesting
+ @Getter(AccessLevel.PROTECTED)
+ private JobState jobState;
+ @VisibleForTesting
+ @Getter(AccessLevel.PROTECTED)
+ private String jobIssuesSummary;
+
Review Comment:
I believe we can directly make them protected instead of private
##########
gobblin-yarn/src/main/java/org/apache/gobblin/yarn/GobblinYarnAppLauncher.java:
##########
@@ -380,6 +382,19 @@ public void launch() throws IOException, YarnException,
InterruptedException {
}, 0, this.appReportIntervalMinutes, TimeUnit.MINUTES);
addServices();
+
+ synchronized (this.applicationDone) {
+ while (!this.applicationCompleted) {
+ try {
+ this.applicationDone.wait();
+ if (this.applicationFailed) {
+ throw new RuntimeException("Gobblin Yarn application failed");
+ }
+ } catch (InterruptedException ie) {
+ LOGGER.error("Interrupted while waiting for the Gobblin Yarn
application to finish", ie);
+ }
+ }
+ }
Review Comment:
Can you please add comment on the importance of this wait ?
##########
gobblin-temporal/src/main/java/org/apache/gobblin/temporal/yarn/YarnService.java:
##########
@@ -353,7 +370,11 @@ protected void shutDown() throws IOException {
}
}
-
this.amrmClientAsync.unregisterApplicationMaster(FinalApplicationStatus.SUCCEEDED,
null, null);
+ if (this.jobState != null && !this.jobState.getState().isSuccess()) {
+
this.amrmClientAsync.unregisterApplicationMaster(FinalApplicationStatus.FAILED,
this.jobIssuesSummary, null);
+ } else {
+
this.amrmClientAsync.unregisterApplicationMaster(FinalApplicationStatus.SUCCEEDED,
null, null);
+ }
Review Comment:
Should we add issuesSummary even in succeed cases as well ?
Issue Time Tracking
-------------------
Worklog Id: (was: 955781)
Time Spent: 40m (was: 0.5h)
> Fail Azkaban job on when temporal job fails
> -------------------------------------------
>
> Key: GOBBLIN-2193
> URL: https://issues.apache.org/jira/browse/GOBBLIN-2193
> Project: Apache Gobblin
> Issue Type: Improvement
> Reporter: Swapnil Palash
> Assignee: Hung Tran
> Priority: Major
> Time Spent: 40m
> Remaining Estimate: 0h
>
> Currently when the temporal job running on Yarn fails, we don't propagate the
> error back to Azkaban job which launches the Yarn Application.
> The change here bubbles the issues encountered when the job fails upto the
> GobblinYarnAppLaucher run by the Azkaban job and fails with a
> RuntimeException after logging the issues summary.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
