Details
-
Suggestion
-
Resolution: Fixed
-
1
-
Description
Problem Definition
In Jira 8.10, automatic stale nodes clean-up solution was presented - JRASERVER-42916.
In some scenarios two days time period to move node from Active, no heartbeat state to the Offline state is too long. This problem becomes more significant in the case of managing the cluster deployed with Ansible, AWS/K8s auto-scaling etc.
Solution
Add system property to change the default retention period for stale nodes - "jira.not.alive.active.nodes.retention.period.in.hours"
Example
Current flow:
ACTIVE → <5 minutes> → NO HEARTBEAT → <2 days> → OFFLINE → <2 days> → GONE
Flow with new system property:
When you will add to your JVM flags on startup this value:
-Djira.not.alive.active.nodes.retention.period.in.hours=3
Then the flow should looks like:
ACTIVE → <5 minutes> → NO HEARTBEAT → <3 hours> → OFFLINE → <2 days> → GONE
Node will be moved to Offline state after 3 hours instead of 2 days.