[GitHub] [zookeeper] stop-coding edited a comment on pull request #1745: ZOOKEEPER-4352:inconsistent data, when minCommittedLog=0, only diff d…

[GitBox <gi...@apache.org>]

stop-coding edited a comment on pull request #1745:
URL: https://github.com/apache/zookeeper/pull/1745#issuecomment-994221309


   > * @stop-coding  Thanks for your contribution.
   > * How can we reproduce this issue?  Do you observe the inconsistent data in the production? Finding a approach to reproduce this issue will give us a deeper understanding for this bug.
   
   Yes, I have meet some times  in scaling up of cluster with k8s.
   The issue is that:
   ###Step 1 -- There is three node (named zk-1,zk-2,zk3) of cluster to scale up. First add zk-1 which would be leader, **_client write some data to zk-1_**. Then add zk-2 to cluster also OK, zk-2 will be follower and sync data from zk-1. 
   ###Step 2 -- Before add zk-3,  zk-1 have connect lost with zk-2, that would make elect leader again, then connection restored and _**zk-2 become leader**_. 
   ###Step 3 -- On this time,  zk-3 joins cluster at the same time, and sync data from zk-2. After that, I find zk-3 only get part of the data form zk-2. Some transaction log which write in zk-1 have lost on zk-3.But the  lost data  exist unexpectedly  in zk-1 and zk-2.
   
   Logs and others on this issues.
   issues link: [https://issues.apache.org/jira/browse/ZOOKEEPER-4352]


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: notifications-unsubscribe@zookeeper.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org