Alphayeeeet
commented
1 month ago Suggesting also somehow, to set a different data view in the built-in dashboards, as metrics-* can be quite large in a fully built production environment. This could be a huge performance improvement for future, so please consider creating a separate issue in the affected projects for this.
Also as an alternative: Consider integrations to create data-views for them, that query only the respecitive module like in this case metrics-kubernetes.*
Integration Name
Kubernetes [kubernetes]
Dataset Name
No response
Integration Version
1.68.1
Agent Version
8.15.2
Agent Output Type
elasticsearch
Elasticsearch Version
8.15.2
OS Version and Architecture
RHEL 8.9
Software/API Version
No response
Error Message
Visualization: error while executing search Control element: Internal Server Error (at least sometimes, as I couldn't reproduce for screenshot).
The Elasticsearch log prints out this message:
Event Original
No response
What did you do?
After some of the previous integration updates (presumably https://github.com/elastic/integrations/pull/10593), the newly added visualizations exceed the bucket limit, and break also the filter control element (sometimes, as I couldn't reproduce when creating the screenshot.). The control element for "Cluster Name" then said: "Internal Server Error.
What did you see?
A uncompletely loaded dashboard like this:
What did you expect to see?
A fully loaded dashboard like this:
Anything else?
Important side node: The time range in all the above screenshots is past 15 min. Longer timeranges not even considered.
Also: Increasing the max_bucket setting helped. But only after increasing it excessively to > 100k and < 1 mio (I didn't try any in between).