Created attachment 1716192 [details] Good display Created attachment 1716192 [details] Good display Description of problem: Graph can't be displayed with perform metric query with big data Version-Release number of selected component (if applicable): 4.6.0-0.nightly-2020-09-23-022756 How reproducible: Always Steps to Reproduce: 1.Login in cluster console as admin, click "Monitoring -> Metrics" 2.Input query "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" 3.click button 'Run Queries' Actual results: Page displays: 1. Ungraphable results 2. The resulting dataset is too large to graph. Expected results: Graph displays well Additional info: 4.6.0-0.nightly-2020-09-21-093308 can display graph, return 729 series 4.6.0-0.nightly-2020-09-23-022756 can't display, return 1004 data series Prometheus UI can display graph
Created attachment 1716193 [details] bad display
When execute query, graph can't be displayed also, but wit different error sort_desc(count({__name__=~"[a-z].*"}) by (__name__)) topk(10,count by (job)({__name__=~".+"}))
Created attachment 1716525 [details] Another error
Closing this as NOTABUG We need to handle the situation where there is too much data to graph in some way. Our current approach is simply to display a message saying there is too much data so the user can adjust the query to return less data. There may be a better approach, but this is currently working as intended. The second error is being returned by Prometheus. The frontend handles Prometheus errors by displaying the error message in the UI, which I think is the right approach since we don't know what the cause of the error is other than by checking the message. (In the case of a timeout, Prometheus returns a 503 response.)