GitHub user jiajunwang opened a pull request:
https://github.com/apache/helix/pull/237
ZkClient related fix and DelayedAutoRebalancer fix.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/jiajunwang/helix master
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/helix/pull/237.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #237
----
commit 41a35eef67737fcbb3f6f97dda7f568a95ef24ed
Author: Jiajun Wang <jjwang@...>
Date: 2018-06-01T17:43:14Z
Increase TestZkConnectionLost timeout to stable the test.
RB=1326718
G=helix-reviewers
A=lxia
commit 5530a7ec59df5af65fecb775f69ee77b6464bbe0
Author: Jiajun Wang <jjwang@...>
Date: 2018-06-01T22:30:39Z
DelayedAutoRebalancer should calculate assignment based on full partition
list.
An issue was report that DelayedAutoRebalancer will generate different
assignment when user change their user-defined preference list. This is because
for some full-auto rebalance strategy, the algorithm balance partition based on
workload. As a result, config change in one partition will affect others.
In this change, we changed the DelayedAutoRebalancer to calculate
assignment based on full partition list first. Then apply user-defined list.
This will fix the issue.
Also updated related test to cover this scenario.
RB=1327182
BUG=HELIX-1052
G=helix-reviewers
A=lxia
commit 9daf026973877158b702652f4c440f46f1bdc075
Author: Jiajun Wang <jjwang@...>
Date: 2018-06-05T17:44:55Z
Make ZkClient keep retrying connect on expiring.
This is to prevent Zk reconnect failure due to transient network issue.
With this change in ZkClient, HelixManager retry is no longer needed.
Deprecate the related option item and simplify handleSessionEstablishmentError
logic.
commit e564744e59ef86f8e45fcf746b7e9b8a3c58c4e4
Author: Jiajun Wang <jjwang@...>
Date: 2018-06-07T00:07:31Z
Fix ZK configurations in the tests.
----
---
