[
https://issues.apache.org/jira/browse/GOBBLIN-1764?focusedWorklogId=839723&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-839723
]
ASF GitHub Bot logged work on GOBBLIN-1764:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 17/Jan/23 17:31
Start Date: 17/Jan/23 17:31
Worklog Time Spent: 10m
Work Description: Will-Lo commented on code in PR #3623:
URL: https://github.com/apache/gobblin/pull/3623#discussion_r1072534871
##########
gobblin-service/src/main/java/org/apache/gobblin/service/monitoring/KafkaJobStatusMonitor.java:
##########
@@ -288,6 +303,15 @@ private static void
modifyStateIfRetryRequired(org.apache.gobblin.configuration.
state.removeProp(TimingEvent.FlowEventConstants.DOES_CANCELED_FLOW_MERIT_RETRY);
}
+ private static boolean
isStateTransitionToFinal(org.apache.gobblin.configuration.State currentState,
List<org.apache.gobblin.configuration.State> prevStates) {
+ Set<String> finalStates = ImmutableSet.of(ExecutionStatus.COMPLETE.name(),
ExecutionStatus.CANCELLED.name(), ExecutionStatus.FAILED.name());
+ if (prevStates.size() == 0) {
+ return
finalStates.contains(currentState.getProp(JobStatusRetriever.EVENT_NAME_FIELD));
+ }
+ return currentState.contains(JobStatusRetriever.EVENT_NAME_FIELD) &&
finalStates.contains(currentState.getProp(JobStatusRetriever.EVENT_NAME_FIELD))
+ &&
!finalStates.contains(prevStates.get(prevStates.size()-1).getProp(JobStatusRetriever.EVENT_NAME_FIELD));
Review Comment:
I think it's impossible for a state to flip flop because of the way how we
perform `mergeState`, which follows an ordered execution status of
```
private static final List<ExecutionStatus> ORDERED_EXECUTION_STATUSES =
ImmutableList
.of(ExecutionStatus.COMPILED, ExecutionStatus.PENDING,
ExecutionStatus.PENDING_RESUME, ExecutionStatus.PENDING_RETRY,
ExecutionStatus.ORCHESTRATED, ExecutionStatus.RUNNING,
ExecutionStatus.COMPLETE,
ExecutionStatus.FAILED, ExecutionStatus.CANCELLED);
```
I think there is a risk if there is a COMPLETE event followed by a FAILED or
CANCELLED event but I believe the first "final" event state we run into is the
one we want users to accept.
Issue Time Tracking
-------------------
Worklog Id: (was: 839723)
Time Spent: 2h 10m (was: 2h)
> Emit GaaSObservabilityEvent
> ---------------------------
>
> Key: GOBBLIN-1764
> URL: https://issues.apache.org/jira/browse/GOBBLIN-1764
> Project: Apache Gobblin
> Issue Type: New Feature
> Components: gobblin-service
> Reporter: William Lo
> Assignee: Abhishek Tiwari
> Priority: Major
> Time Spent: 2h 10m
> Remaining Estimate: 0h
>
> GaaSObservabilityEvents are a new events that provides a job summary from
> pipelines in GaaS. It differs from GobblinTrackingEvents as it runs once per
> job pipeline, and it intended to be easily queryable and alert on.
> We want to emit this observability event from GaaS by deriving it from a
> job's job status. Since this feature is Experimental and WIP, it is not
> expected to fill out all of the fields immediately.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
