ng-monitoring
ng-monitoring copied to clipboard
pingcap
When network connection is re-established (so that local IP is changed), TiDB Dashboard keeps using a wrong advertise address, which makes TiDB Dashboard failed to connect to the ng-monitoring server:...
In manual profiling mutex is provided as a text, but in conprof it is provided as a svg graph.
The current code base is not carefully reviewed. There might be either design problems or logic faults hidden.
Now there is no up-to-date swagger API, so that it is hard for the frontend to keep track with the API interface.
Considering that conprof data is not zstd'ed at all, maybe there is no need any more to split data of different kind into different tables?
Example: ``` ./ng-monitoring-server 2021-11-30 21:10:22.028505 I | Failed to initialize config, err: unexpected empty pd endpoints, please specify at least one, e.g. --pd.endpoints "[127.0.0.1:2379]" ``` Some outputted log format is...
Currently it is not possible to specify different log levels via the command line args, which is not convenient.
Now ngm can conprof other instances, but not itself. Sounds like "医者不能自医".