[ZBX-15036] system.cpu.util[,user] has wrong data on higher thread count systems Created: 2018 Oct 18 Updated: 2019 Feb 18 Resolved: 2019 Feb 18 |
|
Status: | Closed |
Project: | ZABBIX BUGS AND ISSUES |
Component/s: | Templates (T) |
Affects Version/s: | 4.0.0 |
Fix Version/s: | 4.0.2rc1, 4.2.0alpha1, 4.2 (plan) |
Type: | Problem report | Priority: | Major |
Reporter: | Stefan Kocis | Assignee: | Alex Kalimulin |
Resolution: | Fixed | Votes: | 0 |
Labels: | cpu | ||
Remaining Estimate: | Not Specified | ||
Time Spent: | Not Specified | ||
Original Estimate: | Not Specified | ||
Environment: |
Zabbix 4.0.0 LTS Client/target machines cat /proc/cpuinfo (last "core"/thread). |
Attachments: | Screenshot from 2018-11-15 11-29-30.png image-2018-10-18-17-40-15-580.png image-2018-10-18-17-41-07-452.png image-2018-10-18-17-46-52-230.png | ||||||||
Issue Links: |
|
||||||||
Team: | Team A | ||||||||
Sprint: | Sprint 45, Sprint 46, Nov 2018 | ||||||||
Story Points: | 0.5 |
Description |
Steps to reproduce:
I tried 3.0 and 4.0 version of agent with the same result. Other machines that have 40 or less threads work just fine with both centos 6.1 and 7.5 and zabbix agent 2.2 and 3.0 (latest for each distribution). Result:
|
Comments |
Comment by Glebs Ivanovskis [ 2018 Oct 22 ] |
Duplicates ZBXNEXT-391? |
Comment by Stefan Kocis [ 2018 Oct 22 ] |
It does not as far as I can see. But maybe somebody more knowledgeable could provide more certain answer. Edit: on second look it could be connected somehow, but most certainly not the same, since the user load was somewhere between 30% and 50% which would, in the worst case, resulted in ~20% difference and the difference was around 50% and more or less at the level of real user load read from the server as seen on the screenshot.
|
Comment by Stefan Kocis [ 2018 Oct 22 ] |
It seems that this could be it: |
Comment by Stefan Kocis [ 2018 Oct 22 ] |
Just for the reccord. It seems that it is more related to kernel version than CPU cores/threads.
|
Comment by Andris Zeila [ 2018 Oct 26 ] |
It seems that the graphs are missing guest, guest_nice times. It seems you are using virtualization, so that might account for the dips in cpu utilization graphs. Could you try adding them and check? |
Comment by Stefan Kocis [ 2018 Oct 26 ] |
As I wrote earlier. It is exactly that. Thank you. You can probably close this issue if you do not wish to change templates to prevent this kind of misunderstanding. |
Comment by Andris Zeila [ 2018 Oct 29 ] |
Yes, the templates must be updated. |
Comment by Alex Kalimulin [ 2018 Nov 16 ] |
Fixed in:
|
Comment by Alexander Vladishev [ 2018 Nov 20 ] |
(1) [D] Documentation needs to be updated vso we have talked to vzhuravlev and it was not clear how can it be added to official templates. It was suggested that we only change xml templates in svn for now so Zabbix Templates/Official Templates is Won't Fix martins-v This subissue is resolved in (2). WON'T FIX. |