[FLINK-17969] Enhance Flink (Task) logging to include job name as context diagnostic information

• Type: Improvement
• Status: Closed
• Priority: Not a Priority
• Resolution: Won't Fix
• Affects Version/s: 1.10.0
• Component/s: Runtime / Task
• Labels: auto-deprioritized-minor pull-request-available
• Assignee: Unassigned
• Reporter: Bhagavan
• Created: 27/May/20 12:24
• Updated: 15/Sep/22 11:38
• Resolved: 15/Sep/22 11:38

Description

Problem statement:
We use a shared session cluster (Standalone/Yarn) to execute jobs. All logs from the cluster are shipped using log aggregation framework (Logstash/Splunk) so that application diagnostic is easier.
However, we are missing one vital information in the logline. i.e. Job name so that we can filter the logs for a single job.

Background
Currently, Flink logging uses SLF4J as API to abstract away from concrete logging implementation (log4j 1.x, Logback or log4j2) and configuration of logging pattern and implementation can be configured at deployment, However, there is no MDC info from framework indicating job context.

Proposed improvement.

Add jobName field to Task class so that we can add it as MDC when task thread starts executing.

Change is trivial and uses SLF4J MDC API.

With this change, user can customise logging pattern to include MDC (e.g. in Logback %X

{jobName}

)

Change required.

@@ -319,6 +323,7 @@ public class Task implements Runnable, TaskSlotPayload, TaskActions, PartitionPr
 
                this.jobId = jobInformation.getJobId();
+               this.jobName = jobInformation.getJobName();
                this.vertexId = taskInformation.getJobVertexId();
@@ -530,8 +535,10 @@ public class Task implements Runnable, TaskSlotPayload, TaskActions, PartitionPr
        @Override
        public void run() {
                try {
+                       MDC.put("jobName", this.jobName);
                        doRun();
                } finally {
+                       MDC.remove("jobName");
                        terminationFuture.complete(executionState);
                }
        }

if we are in agreement for this small change. Will raise PR.

Issue Links

• links to: GitHub Pull Request #12472