This helps with orderly Overlord failover. After this patch, by the time the task runner returns from start(), all tasks are located (subject to some timeout). This is useful for supervisors, which start next and which need to contact tasks.
The timeout defaults to 1 minute, and is configurable using druid.indexer.runner.taskJoinTimeout.
This helps with orderly Overlord failover. After this patch, by the time the task runner returns from
start(), all tasks are located (subject to some timeout). This is useful for supervisors, which start next and which need to contact tasks.The timeout defaults to 1 minute, and is configurable using
druid.indexer.runner.taskJoinTimeout.