[HDDS-3580] Retry Cache in OM

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.1.0
• Component/s: OM HA
• Target Version/s: 1.1.0
• Assignee: Bharat Viswanadham
• Reporter: Bharat Viswanadham
• Created: 13/May/20 19:23
• Updated: 08/Dec/20 18:24
• Resolved: 08/Dec/20 18:24

Description

The failover proxy provider for OM HA retries requests on Network Timeout/Connectivity Exception. It can retry the request on the same OM or a new OM. If the request has been already processed, the subsequent retry will fail with error for non idempotent operations such as createVolume, createBucket createKey, renameKey etc. This will cause an application failure, unnecessary overhead in processing the operation again, and possibly corruption in metadata during failover.

This Jira is to discuss this issue in detail, and also categorize OM operations.

Sub-Tasks

• Use ClientID and CallID from Rpc Client to detect retry requests: Resolved
• Expose Ratis retry config cache in OM: Resolved