TiCDC sink re-establishes in some unnecessary cases

[hicqu]

What did you do?

After TiCDC sink is stuck for a while, TiCDC will destroy and re-establish the sink. However currently the stuck detection is not good enough. In such cases stuck will be reported mistakely:

• after a long time incremental scan;
• after a sync point is passed;
• if a table hasn't been scheduled for a while.

What did you expect to see?

No response

What did you see instead?

In those cases, sink will be re-established, which is not expected.

Versions of the cluster

Master branch.

Analysis

Users who configured consistent.level = eventual and enable-syncpoint = true can meet the problem more easily. Especially for the cases with syncpoint-interval is larger than 150s, and it's the most common situations in fact.

It will cause table sink re-establishing too frequently, even cause checkpoint can not be advanced at all.