New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Dashboard search performance regressed after upgraded to v9 #55332
Comments
Try enabling our new search engine [feature_toggles]
panelTitleSearch = true |
Unfortunately, that is not an option for us. We tried to enable this feature toggle in our staging environment and Grafana just went into Crash loops with this error.
Just to confirm, without enabling the |
Correct. How many dashboards? Do you have many LARGE dashboards (perhaps snapshots?) |
We have 12K dashboards. How do you define LARGE dashboards? |
Hey @yunkaisc! First of all, thank you for your patience and for reporting this issue - we are still working on the new search, and are actively trying to make it better. We have recently added a few configuration options that might help you with this problem. They are available in both 9.1.7 (https://github.com/grafana/grafana/blob/v9.1.x/conf/defaults.ini#L1289-L1300) and 9.2.0 releases. Would you be willing to try the following config
and letting us know if this issue still persists? |
I have taken a similar search performance hit after upgrading from 8.5.3 to 9.2.4. I have 19,112 dashboards. I turned on panelTitleSearch feature toggle but it doesn't seem to have helped the performance. If I set full_reindex_interval=15m does that extend the context deadline mentioned above to 15 minutes? |
I believe I'm also seeing this issue upgrading from 8.3.5 to 9.2.4. We have 15k dashboards and are using a Postgres DB. We saw large increases in Postgres CPU usage. I looked at the slow DB queries and queries like the one below were taking up like 98% of DB CPU time
|
scaling up Postgres 4x improved our search performance 4x |
@frankyi-gh when you say "scaling up Postgres 4x" do you mean 4x CPU, 4x memory, or something else? If 4x CPU, did you go from 2 CPUs to 8 CPUs? I'm just trying to get a feel for magnitude of your resource increase. |
4x CPU and 4x memory. Specifically, I went from an AWS db.r5.xlarge to db.r5.4xlarge |
Turning on panelTitleSearch with the default search settings seems to have actually increased DB load, which made general performance a little bit worse. It didn't do much for search times. |
Thank you, I appreciate the tuning details. I will say that I really do like the new panelTitleSearch feature as I have automation in place that stores a wealth of information (circuit ids, physical node names, etc.) in panel titles. It's really nice to be able to use Grafana's search to dig through that data. |
I've been told that this issue will be fixed by #56813 in 9.3 |
9.3 with panelTitleSearch enabled doesn't seem any better... |
Hello, updating from grafana-8.5.5-1 to grafana-9.3.2-1 brought me to this case. We also are facing the same issue but what can I say is that the problem of slow dashboard lists impact only non-admin users.
is helping me at the moment. The mariadb slow query log pointed us to the following query:
If in that query we remove the:
the query run super fast again. |
Hello! Tested now with 9.4.2 and the dashboard list is OK, run very well. Thanks |
Thanks for confirming that the fix has improved the search performance @afeshti! I'll close this issue now 🎉 |
What happened:
After we upgraded Grafana from v8.5.4 to v9.1.4, we noticed a significant performance regression when searching the dashboard.
Based on the data points from logs,
In v8.5.4, the duration of a dashboard search request is much shorter. The percentiles are based on 2,500+ searches.
In V9.1.5, the duration of a dashboard search request is noticeably longer. The percentiles are based on 3,000+ searches.
What you expected to happen:
Expect the dashboard search is as fast as v8.
How to reproduce it (as minimally and precisely as possible):
Noticeably slower than before. Also
Anything else we need to know?:
Environment:
The text was updated successfully, but these errors were encountered: