elastic / kibana

Your window into the Elastic Stack
https://www.elastic.co/products/kibana
Other
19.33k stars 7.97k forks source link

Process tab in Host monitoring seems to be confusing users, report `N/A` when data exists #181914

Open fearful-symmetry opened 1 week ago

fearful-symmetry commented 1 week ago

So, the the Process tab in host monitoring presents a summary of process state at the top:

image

It appears that this UI has spawned an entire issue from users who are concerned that something in beats is broken: https://github.com/elastic/beats/issues/38120

The problem is that on most linux systems, the majority of processes will report as a sleep state unless are under heavy CPU load. Users who aren't experienced with linux process scheduling, see something counter-intuitive like "everything on the host is sleeping" and assume there's a problem.

There's also an adjacent problem, demonstrated in the screenshot, where the process states show N/A despite the process list showing a number of running processes.

I'm not sure what to do here. Part of me thinks we should just remove the process state summary element at the top, as for most use cases, counts of running processes aren't particularly helpful. We could also replace it with a count of total processes, or provide some kind of tooltip.

elasticmachine commented 1 week ago

Pinging @elastic/obs-ux-infra_services-team (Team:obs-ux-infra_services)

crespocarlos commented 1 week ago

There's also an adjacent problem, demonstrated in the screenshot, where the process states show N/A despite the process list showing a number of running processes.

@fearful-symmetry , the reason for that is because the process_summary metricset needs to be enabled in the system module integration. Perhaps we could replace the need for that metricset by using the information provided on the list, but I'm not sure if we'll have all the information that the summary shows.

fearful-symmetry commented 1 week ago

@crespocarlos yeah, the link between that particular visualization and the process_summary metricset is a bit non-obvious, even I assumed it was just aggregating the data from the below list. The same data is present a few inches away, so it makes it look like a bug.

roshan-elastic commented 1 week ago

Hey @fearful-symmetry - following on from my previous comment (on the related issue)...would it help to have a call to talk this through?

fearful-symmetry commented 1 week ago

@roshan-elastic I wouldn't be opposed to that, but be warned, I have a lot of strong opinions about system metrics UX ;)

roshan-elastic commented 1 week ago

Thanks @fearful-symmetry

I don't have strong opinions at this stage :)

I'll set something up.