cdc maybe stuck when send SIGSTOP to pd leader

[zeminzhou]

What did you do?

1. tiup start one cluster with two cdc & three cdc
2. kill -19 to pd leader
3. cdc cli capture list can't list all capture, if cdc cli capture list sucessfully, repeat step 2.

What did you expect to see?

cdc cli capture list list all capture

What did you see instead?

cdc cli capture list can't list all capture

Versions of the cluster

Upstream TiDB cluster version (execute SELECT tidb_version(); in a MySQL client):

v6.3.0

Upstream TiKV version (execute tikv-server --version):

v6.3.0 

TiCDC version (execute cdc version):

v6.3.0

[zeminzhou]

From debug, capture routine may exit when kill -19 to pd leader, and capture routine will recover by itself. Then capture routine will call reset, reset will call concurrency.NewSession, capture routine stuck here.

[zeminzhou]

When kill -19 to pd leader, capture(that is not pd leader )'s processor routine will exit and the owner routine will struck(even if cancel is called). From debug, owner routine will stuck here.