http api “api/v1/changefeeds” was refused by owner when cdc cluster is rolling update

[maxshuang]

What did you do?

1. Update cdc version one by one in cdc cluster
2. Query changefeed status in the middle of rolling update:
   curl -X GET http://172.16.6.140:8300/api/v1/changefeeds
   （owner is 5.3.0-master，other is 5.2.1）

What did you expect to see?

query successfully on owner

What did you see instead?

[maxshuang@copy-of-vm-ee-centos76-v1 tidb_cluster]$ curl -X GET http://172.16.6.140:8300/api/v1/changefeeds
{
    "error_msg": "context deadline exceeded",
    "error_code": ""
}

[maxshuang@copy-of-vm-ee-centos76-v1 tidb_cluster]$ curl -X GET http://172.16.6.140:8301/api/v1/changefeeds
[
    {
        "id": "kv-reconn",
        "state": "normal",
        "checkpoint_tso": 429141621833990146,
        "checkpoint_time": "2021-11-16 14:49:31.376",
        "error": {
            "addr": "172.16.6.140:8300",
            "code": "CDC:ErrOwnerUnknown",
            "message": "[CDC:ErrReachMaxTry]reach maximum try: 20: [CDC:ErrMySQLTxnError]Error 1049: Unknown database 'workload': Error 1049: Unknown database 'workload'"
        }
    }

It seems that owner refuses to serve the http request. Is it acceptable?

Versions of the cluster

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

5.2.1

TiCDC version (execute cdc version):

5.3.0-master