[ZOOKEEPER-4289] Reduce the performance impact of Prometheus metrics

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.6.3, 3.7.0, 3.6.2, 3.8.0, 3.7.1
• Fix Version/s: 3.9.0, 3.8.1
• Component/s: metric system
• Labels: pull-request-available
• Assignee: Li Wang
• Reporter: Li Wang
• Created: 04/May/21 01:13
• Updated: 30/Jan/23 07:57
• Resolved: 23/Jan/23 13:07

Description

I have done load comparison tests for Prometheus enabled vs disabled and found the performance is reduced about 40%-60% for both read and write operations (i.e. getData, getChildren and createNode).

Looked more into this and found that Prometheus Summary is very expensive and there are more than 20 Summary metrics added to the new CommitProcessor.

Need a solution to reduce the impact of Prometheus metrics.

Issue Links

• links to: GitHub Pull Request #1698