Use more retries in tree pool in case of failures #172

New issue

Closed

opened 2023-10-05 11:48:11 +00:00 by alexvanin · 1 comment

alexvanin commented

2023-10-05 11:48:11 +00:00

Owner

There is a case, when put request to tree service happens on one node and get request happens on another node in the same container. Tree replication can not be immediate, so temporary client returns empty result or even Tree Not Found error.

Pool should try its best to receive node for tree services, so it's better to repeat request for different endpoints in this case.

Describe the solution you'd like

Use more retries in case of bad and unexpected responses.

Describe alternatives you've considered

Offload retries to storage node.

Additional context

This was highlighted during failover tests with nginx as load-balancer and some sort of round-robin

## Is your feature request related to a problem? Please describe. There is a case, when put request to tree service happens on one node and get request happens on another node in the same container. Tree replication can not be immediate, so temporary client returns empty result or even `Tree Not Found` error. Pool should try its best to receive node for tree services, so it's better to repeat request for different endpoints in this case. ## Describe the solution you'd like Use more retries in case of bad and unexpected responses. ## Describe alternatives you've considered Offload retries to storage node. ## Additional context This was highlighted during failover tests with nginx as load-balancer and some sort of round-robin