[AMBARI-18271] Multiple Aggregate Alerts For JournalNode Exist With Some Being Incorrect

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 2.0.0
• Fix Version/s: trunk, 2.4.1
• Component/s: ambari-server
• Assignee: Jonathan Hurley
• Reporter: Jonathan Hurley
• Created: 29/Aug/16 16:29
• Updated: 08/Sep/16 18:36
• Resolved: 30/Aug/16 03:37

Description

When a cluster is being created initially, there are many new alert instances being generated concurrently. This can lead to a race condition where certain alert types, such as aggregate alerts, fire several events and end up creating duplicate alert instances.

Although aggregate alerts are the most common, the flaw is in the alert listener since it never guarantees uniqueness even though it supports concurrent events.

This problem is mainly encountered when the cluster is initially created, and therefore a solution shouldn't impact the rest of the cluster's lifetime. In other words, we shouldn't cause a degradation the other 99.9% of the time in order to fix this problem.

Issue Links

• links to: Reviewboard