changefeed get stuck in owner switch case

[Tammyxia]

What did you do?

• 5 captures, 1 changefeed, downstream is mysql
• run CDC owner switch case in test infra, ( tpcc run parallelize with owner switch)

What did you expect to see?

no error

What did you see instead?

• changefeed get stuck
• changefeed can be paused or resumed

Versions of the cluster

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

5.0.6

TiCDC version (execute cdc version):

Release Version: v5.0.6
Git Commit Hash: eba76b8952647d1cb648031f0f505e4d7ff91acf
Git Branch: heads/refs/tags/v5.0.6
UTC Build Time: 2021-12-23 12:26:43
Go Version: go version go1.13 linux/amd64
Failpoint Build: false

[Tammyxia]

goroutines.log
ticdc.log

[Tammyxia]

cdc log:
[2021/12/23 18:32:06.369 +00:00] [WARN] [client.go:1157] ["failed to receive from stream"] [addr=upstream-tikv-0.upstream-tikv-peer.cdc-testbed--tps-512789-1-548.svc:20160] [storeID=1] [error="rpc error:
[2021/12/23 18:32:06.369 +00:00] [WARN] [mysql.go:961] ["failed to rollback txn"] [error="context canceled"]
[2021/12/23 18:32:06.369 +00:00] [WARN] [mysql.go:961] ["failed to rollback txn"] [error="context canceled"]
[2021/12/23 18:32:06.369 +00:00] [ERROR] [mysql.go:1104] ["execute DMLs failed"] [err="[CDC:ErrMySQLTxnError]context canceled"]
[2021/12/23 18:32:06.369 +00:00] [ERROR] [mysql.go:1104] ["execute DMLs failed"] [err="[CDC:ErrMySQLTxnError]context canceled"]

[Tammyxia]

ticdc0.log