You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@samza.apache.org by "Sanil Jain (Jira)" <ji...@apache.org> on 2019/10/24 00:11:00 UTC

[jira] [Created] (SAMZA-2361) Revert Request Expiry check for Host affinity disabled cases

Sanil Jain created SAMZA-2361:
---------------------------------

             Summary: Revert Request Expiry check for Host affinity disabled cases 
                 Key: SAMZA-2361
                 URL: https://issues.apache.org/jira/browse/SAMZA-2361
             Project: Samza
          Issue Type: Bug
            Reporter: Sanil Jain
            Assignee: Sanil Jain


**Symptom: **: PR #1170 introduced expired request checking for Host Affinity disabled cases, however, this leads to a few problems that we might face in future:
1. In the scenarios of a cluster bounce, this might result in jobs (in either case of host affinity enabled & disabled cases) requesting containers from the ClusterResourceManager every 5 (default request expiry timeout) seconds if requests can't be fulfilled. Before PR #1170 this only used to happen for host affinity disabled jobs. This might potentially result in bringing down the ClusterManager. 
2. Unless we have some performance benchmarking on RM (for Yarn as a cluster balancer), having a default expiry timeout of 5 seconds for all container requests is not justified.

**Cause:** PR #1170

**Fix:** We are reverting this behavior change since we have Performance benchmarking for Yarn

**Tests:** N/A



--
This message was sent by Atlassian Jira
(v8.3.4#803005)