The metric `Kafka Outgoing Bytes` is inaccurate.

[asddongmen]

What did you do?

The lag of this monitoring indicator is significant. For instance, if the original traffic was 100MB and the actual traffic dropped tenfold, this indicator would not immediately reflect the current traffic. Instead, it could take almost 2 hours to update.

What did you expect to see?

It can reflect the current traffic immediately.

What did you see instead?

Described above.

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):

Master.