Metricbeat Default Dashboards TSVB and Histogram Display Issues after Upgrade from 7.17 to 8.7

Hello,

I recently upgraded my Elastic Stack from version 7.17 to 8.7, and I am experiencing issues with the default Metricbeat dashboards. The TSVB and Histogram visualizations are not displaying data correctly. I have observed that if I change the 'top' value (number of hosts to display) to a number close to the actual number of servers, the data is shown correctly. However, when the 'top' value is set to the default 10, the data is displayed as shown in the attached image.



This issue specifically affects the 'average' metric. When I change the metric to 'max,' the data is displayed correctly. I have attached screenshots for your reference.

Additionally, I have noticed that almost all of my dashboards have lost their URLs after the upgrade. Is there a way to reverse this issue or recover the original URLs?

Please advise on how to resolve these issues, as I am unable to use the default Metricbeat dashboards effectively due to these problems.

Thank you for your assistance.

Best regards

Hi @Adriann !

I'm sorry that dashboard is giving you some issues. It sounds like one problem is that the terms you're seeing from Elasticsearch aren't always the true top 10 values. Did I understand that right?

If so, could you try clicking that "Edit visualization in Lens" button to open this vis in Lens? Once you've done that, you can enable accuracy mode (see the video on the PR) which should improve the accuracy of your results at the expense of some extra load on Elasticsearch.

If you want to understand more about why you don't always get the true top values in this case, my colleague wrote a great explanation of how the underlying tech works. Of course, let me know if you have any questions.

As far as the dashboard URL thing—could you explain a bit more how you're trying to use the URLs? Are you trying to share the dashboard externally?

Hello Andrew_Tate

Thank you for your response. After reading It I thought that probably I did not describe the issue well enough. Basically, the problem arise after I upgraded Stack from 7.17 to 8.7 on the default dashboard. So far I have noticed It only on the two dashboards visualization.
1.) "Top Hosts By CPU (Realtime) [Metricbeat System] ECS" of " [Metricbeat System] Overview ECS"
2.) "Hosts histogram by CPU usage [Metricbeat System] ECS" of " [Metricbeat System] Overview ECS"
3.) "Top Hosts By Memory (Realtime) [Metricbeat System] ECS" of " [Metricbeat System] Overview ECS"

This is what I see for the default dashboard

This is how it looks like when I change the "top" value

There is a lot of empty value from the top of the visualization. I have checked and if I take some name of the server where there is no value in the current visualization. Usually, CPU usage is very low like 0.3%, etc.

When I was writing this post response I notice that when I try to convert to the lens, on the lens page Kibana asks to "Increase Accuracy" and after that, the visualization started to look like before the upgrade. Indeed without color coding but there where some information.

I hope this additional information clarifies the issue I'm facing. Please let me know if you need any further details.

Thank you.

So it sounds like the issue is that you're getting some empty bars as your "top values." Could Terms agg: sort null buckets below numeric buckets · Issue #95239 · elastic/elasticsearch · GitHub be the issue? You can check the Elasticsearch response to make sure by clicking "Inspect."

I'm not sure why this would have changed after upgrade, but if Lens works for you, does that solve your issue?

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.