Hey folks, we recently migrated our grafana stack from 9.x to 12.0. While all seems good and users really like the new UI and features, they complained that large dashboards were somewhat slow. On digging, we found that grafana ui was now sending queries (prometheus datasource) to the backend only in batches of 5-6, i.e at any point in time there were only 5-6 pending requests to the backend. And so, if the panels have a total of 20 queries, they would be executed in batches of 4. Is there a setting to increase the concurrency limit?We tried disabling dashboardScene and setting concurrent_query_limit to 20 but that didnt help.
[feature_toggles] dashgpt = false dashboardScene = false dashboardSceneForViewers = false dashboardSceneSolo = false publicDashboardsScene = false [query] concurrent_query_limit = 20
Our next suspect is http 1.1, which limits max connections per host to 6. Bear in mind that with the recent upgrade, all our datasources converted from direct mode to proxy mode.Has anyone else investigated something similar? I found this thread in the community but there was no conclusion.
https://community.grafana.com/t/setting-number-of-concurrent-queries-or-number-of-repeated-panels-to-render-at-once/120789