|
@@ -0,0 +1,13 @@
|
|
|
|
|
+Improve `/api/v5/monitor` endpoint performance.
|
|
|
|
|
+
|
|
|
|
|
+Prior to this enhancement, the dashboard monitor page often timeout when there are many nodes in the cluster.
|
|
|
|
|
+Below are detailed changes:
|
|
|
|
|
+
|
|
|
|
|
+- Make concurrent RPC calls towards nodes in the cluster to retrieve metrics.
|
|
|
|
|
+- Data points are downsampled to reduce dencity.
|
|
|
|
|
+ Dencity is adjusted according to the requested time span in the query:
|
|
|
|
|
+ - `10s` when query the last `1h`
|
|
|
|
|
+ - `1m` when query the last `1d`
|
|
|
|
|
+ - `5m` when query the last `3d`
|
|
|
|
|
+ - `10m` when query the last `7d`
|
|
|
|
|
+- Insert dummy data points for gaps in the timeline (if EMQX has stopped for a while), so the gap will be visible in the dashboard.
|
zmstone