improper metric usage batch_row_seq_scan_next_duration

[cloudcarver]

Seems batch_row_seq_scan_next_duration stores query_id in the label. This means every query will produce a new time series, this is a bad practice when using metric.

IMHO, the better idea is to use log to record this information. In this case, we can check the exact query id in the log when we find a surge in the abnormal metric graph.

[fuyufjh]

@liurenjie1024 Please take a look, thanks!

[liurenjie1024]

It may cause high cardinality metrics in highly frequent distributed query, but our system assumes that no frequent distributed query happens, so I think it's ok to work. Metrics are disabled for high qps serving queries.

[github-actions]

This issue has been open for 60 days with no activity. Could you please update the status? Feel free to continue discussion or close as not planned.