Bug 1751541 - In the nodes' dashboards, the CPU usage graph is in the wrong units
Summary: In the nodes' dashboards, the CPU usage graph is in the wrong units
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: OpenShift Container Platform
Classification: Red Hat
Component: Console Metal3 Plugin
Version: 4.2.0
Hardware: Unspecified
OS: Unspecified
unspecified
medium
Target Milestone: ---
: 4.3.0
Assignee: Honza Pokorny
QA Contact: Udi Kalifon
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2019-09-12 06:39 UTC by Udi Kalifon
Modified: 2020-01-14 11:48 UTC (History)
6 users (show)

Fixed In Version:
Doc Type: No Doc Update
Doc Text:
Clone Of:
Environment:
Last Closed: 2019-11-13 13:49:11 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)
CPU Usage graph (59.76 KB, image/png)
2019-09-12 07:16 UTC, Udi Kalifon
no flags Details


Links
System ID Private Priority Status Summary Last Updated
Github openshift console pull 2700 0 'None' 'closed' 'Bug 1751541: [metal3] Change cpu utilization to cores' 2019-11-11 13:16:39 UTC

Description Udi Kalifon 2019-09-12 06:39:54 UTC
Description of problem:
When viewing the node's CPU usage, the y axis is supposed to have units of percentage, but instead it's showing seconds.

When I looked at the graph when the node was idle, the y axis showed "500m" and "1", instead of "0.5%" and "1%". By comparison, the same graph in the bare metal host dashboard is showing the results correctly in percentages.


Version-Release number of selected component (if applicable):
4.2.0-0.nightly-2019-09-11-032939


How reproducible:
100%


Steps to Reproduce:
1. Browse to Compute -> Nodes
2. Click on one of the nodes
3. See the CPU usage graph in the node's dashboard


Actual results:
The units on the y axis are apparently seconds instead of percentages

Comment 1 Udi Kalifon 2019-09-12 07:16:15 UTC
Created attachment 1614329 [details]
CPU Usage graph

See attached screen shot of CPU Usage graph.

Comment 2 Jiri Tomasek 2019-09-12 09:14:57 UTC
It actually seems that the unit is correct in Nodes graph and is incorrect in baremetal host cpu utilisation graph. `instance:node_cpu:rate:sum` should use 'humanizeCpuCores' funnction to get correct units.

Comment 3 Udi Kalifon 2019-09-12 12:22:46 UTC
If I check on the node with "top" - I can see the CPU is measured in us. However, when I load the CPU and the usage climbs to ~7.6, the graphs show ~2.5. So the question is really what's the intended units of these graphs, and how to test the result for correctness. By the way to load the CPU - I run in the background: dd if=/dev/zero of=/dev/null &

Comment 4 Honza Pokorny 2019-09-12 13:25:27 UTC
The newly redesigned utilization card for the baremetal host has percentages.

https://docs.google.com/document/d/1q0DIcUKgFhvNbtGt-SMl35Z2jHA_QO2r7aaBuFmwE2U/edit#

Comment 6 Udi Kalifon 2019-09-17 06:51:35 UTC
We're expecting to see percentages, like in the design doc. There are now no units at all - just a plain number.

Comment 7 Honza Pokorny 2019-09-17 11:34:47 UTC
The PR linked changes the units on the CPU graph in the baremetal dashboard to be consistent with the one on the Node detail page.  As such, this shouldn't fail QA.  The percentages that you mention are part of the redesign of the utilization card, and this hasn't been implemented yet.

Comment 8 Wenjing Zheng 2019-10-11 07:45:30 UTC
(In reply to Udi from comment #6)
> We're expecting to see percentages, like in the design doc. There are now no
> units at all - just a plain number.

Udi, per comment #7, could you re-verify this bug?

Comment 11 Samuel Padgett 2019-11-12 15:05:15 UTC
I don't think we use should show percentages. I'd expect the values to be in cores and millicores.

Comment 12 Jiri Tomasek 2019-11-13 13:49:11 UTC
Bare Metal host CPU utilisation is consistent with other CPU utilisation graphs in other dashboards and uses cores/millicores (as described in https://bugzilla.redhat.com/show_bug.cgi?id=1751541#c7 and https://bugzilla.redhat.com/show_bug.cgi?id=1751541#c11). Closing. Feel free to reopen if needed.


Note You need to log in before you can comment on or make changes to this bug.