[HDDS-3046] Fix Retry handling in ozone RPC Client

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.5.0
• Labels: OMHA OMHATest pull-request-available
• Assignee: Bharat Viswanadham
• Reporter: Bharat Viswanadham
• Created: 20/Feb/20 20:34
• Updated: 27/Feb/20 21:48
• Resolved: 24/Feb/20 21:50

Description

Right now for all other exceptions other than serviceException we use FailOverOnNetworkException.

This Exception policy is created with 15 max fail overs and 15 retries. 

 

retryPolicyOnNetworkException.shouldRetry(
 exception, retries, failovers, isIdempotentOrAtMostOnce);

2 issues with this:

1. When shouldRetry returns action FAILOVER_AND_RETRY, it will stuck with same OM, and does not perform failover to next OM.  As OMFailoverProxyProvider#performFailover() is a dummy call does not perform any failover.
2. When ozone.client.failover.max.attempts is set to 15, now with 2 policies with each set to 15, we will retry 15*2 times in worst scenario. 

 

 

Issue Links

• links to: GitHub Pull Request #575