Some regions failed to be deregistered after unregistering the subscription

[lidezhu]

What did you do?

Just as title shows, some regions failed to be deregistered after unregistering the subscription.

As shown in the monitoring screenshot, after the scheduling around 00:00, the number of resolved ts events received by CDC per second spiked from 150K to 160K. This metric returned to the original 150K level after a CDC restart. Therefore, it can be inferred that some regions were not successfully cleaned up after the scheduling operation, causing this residue.

Furthermore, logs containing messages like "subscription client holes exist," which originates from the unregister subscription operation, further corroborate the conclusion above.

What did you expect to see?

No response

What did you see instead?

None

Versions of the cluster

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

(paste TiDB cluster version here)

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

(paste TiKV version here)

TiCDC version (execute cdc version):

(paste TiCDC version here)