Description of problem: ----------------------- I have generated daily, weekly and monthly Chargeback reports on Dec 11 at midnight UTC. period = 12/07 - 12/10 The resource usage is different in all the 3 reports. For eg, in the cb-owner-daily - Sun, 11 Dec 2016 00:00:14 +0000 report,I summed up the CPU used for the cu-24x7 VM for 12/07 - 12/10. Total CPU used = (4.17 + 47.52 + 58.37 + 53.17)MHz = 163.23 MH The weekly report shows that the CPU used is 23.65 MHz for the same period. The monthly report shows that CPU used is 5.27 MHz for the same period. I have attached all the three reports. Version-Release number of selected component (if applicable): ------------------------------------------------------------- 5.7 How reproducible: ----------------- Always Steps to Reproduce: ------------------- 1.Manage a provider. Collect C&U data for a few days(4 days in the attached report). 2.Generate daily Chargeback report with this option : In the filter tab, in the Chargeback interval section,select 'Day' in the 'Show costs by' option in the Chargeback Interval section. Generate weekly Chargeback report with this option : In the filter tab, in the Chargeback interval section,select 'Week' in the 'Show costs by' option Chargeback Interval section. Generate monthly Chargeback report with this option : In the filter tab, in the Chargeback interval section,select 'Month' in the 'Show costs by' option in the Chargeback Interval section. 3.Compare the resource usage reported in the three reports. Actual results: --------------- The resource usage in the weekly and monthly reports are incorrect. Expected results: ----------------- The resource usage in the weekly and monthly reports should be the same as the resource usage reported in the daily report. Additional info: ----------------
Created attachment 1240409 [details] Daily Chargeback report
Created attachment 1240411 [details] Weekly Chargeback report
Created attachment 1240412 [details] Monthly Chargeback report
Hello Nandini, This is very nice find. I noticed this also when working on bug 1402072. It should be fixed together with bug 1402072. The problem was: --------------- We took consumption from metric rollups as A. We took number of hours in interval (day = 24, week=24*7, month ~~ 24*30) as B. Then we calculated average as = A/B. The problem was that we cannot use B as it is. We need to use only past hours in given interval. More Issue like this: --------------- Imagine you create VM in the middle of month. Then consumption average should be again calculated based on real consumed hours. Not hours in month. Imagine you retire VM in the middle of month. Then consumption average should be again calculated based on real consumed hours. Not hours in month. That's all fixed in bug 1402072. Thanks!
Verified in 5.7.1
I meant to verify the 5.7.1 clone and not this. Moving this back to ON_QA.
This issue affects 5.6.4 as well.Can this be cloned to 5.6?
I think the answer is yes. This can be cloned. You can use clone button to make the clone for 5.6. Then, the 3 ack process will be used to decide whether that bug is worth including into the 5.6. I have no visibility into how many releases are planned for 5.6.z. and I am not sure if there are any written policies for backporting to 5.6. Thus, it is still possible that the clone will never be fixed. However, it is worth filing, as it gives us visibility into what is broken in what versions. Thanks!
Verified in 5.8.0.10