The shards which assigned to non-leader instances may not be executed in specific scenario

[TeslaCN]

Bug Report

This problem can be reproduced in specific scenario.

Which version of ElasticJob did you use?

3.0.0-beta-SNAPSHOT

Which project did you use? ElasticJob-Lite or ElasticJob-Cloud?

ElasticJob-Lite

Expected behavior

Actual behavior

1. Leader instance has already finished sharding and started executing shards assigned to it.
2. Non-leader instance's listener set resharding necessary flag before it start executing LiteJob#execute().
3. So the Non-leader instances keep waiting for sharding completed.

Reason analyze (If you can)

Example codes for reproduce this issue (such as a github link).

https://github.com/TeslaCN/shardingsphere-elasticjob-lite/tree/reproduce-blocking/examples/elasticjob-example-lite-java/src/main/java/org/apache/shardingsphere/elasticjob/lite/example/bug

Steps to reproduce the behavior.

Check out my branch and configure zookeeper. Main class is org.apache.shardingsphere.elasticjob.lite.example.bug.ReproduceMain

1. Start a process "A" and wait it schedule job.
2. After process "A" completed schedule, start a new process "B".

[TeslaCN]

I have analyzed logs and draw a graph.

[TeslaCN]

I recorded the process of troubleshooting on my blog.
https://blog.csdn.net/wu_weijie/article/details/108678606