Ling Jin
Ling Jin
The root cause of the issue comes from the kv client cannot unregister the event handle from the worker pool immediately. When close the puller, kv client also close, it...
After analyze the logs, we think this is caused by the flashback command, will be fixed by fix the flashback.
This problem is caused by that one lock is comes from the range which not subscribed by the TiCDC.
If the encoding protocol would send checkpoint to the kafka server, there is no need to send heartbeat.
This is not a bug, can be tagged as `enhancement`.
This is caused by the message duplication, no message loss, this message can be ignored and does not affect eventual consistency. We are going to investigate this issue further
This issue cannot be reproduced, close it. If reproduced, reopen this issue