How can I create a expression to calculate the aggregated past 30 day's SLO consumption

[DylanLi]

Recently, my team successfully implemented SLO monitoring in Dynatrace. We now need to assess the total error budget consumption over the past 30 days basing on various time points to adjust the SLA and refine our alerting strategy. I have observed that Dynatrace can provide aggregated data on the SLO page as follows:

I know that I can change the time frame setting on the right top in this page to get the total 30-day error budget consumption basing on different time point. But I can't figure out a way to use an expression in the Data Explore to get some Line Graph of the 30-day total consumption to show the trends. The expression given in the SLO detailed page can only let me get the error budget consumption in every time sample instead of an aggregated value. 

(100)*(1-(builtin:service.errors.fivexx.count:splitBy())/(builtin:service.requestCount.total:splitBy()))

Can anyone advise on which functions or operators can be used to create a line graph illustrating the 30-day error budget total consumption trends similar to the mock graph provided? For example, the final SLO from February 8th to March 10th is 100%, from February 9th to March 11th is 99.1%, and from February 10th to March 12th is 99.0%.

[Julius_Loman]

Not sure if I got your question right, but probably just setting the resolution to one day will give you the results you need. Also just for the reference , the classic SLOs which are you using, create functions functions for error budgets.

[DylanLi]

Hello Julius, thank you for your response. However, setting the resolution to one day does not seem to fulfill the requirement. With the expression mentioned above,

(100)*(1-(builtin:service.errors.fivexx.count:splitBy())/(builtin:service.requestCount.total:splitBy()))

if I configured the resolution to one day and the timeframe to 30 days, in the resulting line graph, each data point only reflects the error budget consumption for the previous 24 hours. For instance, if the data point for March 11th shows 97.5%, it indicates that the success rate for requests from March 10th 00:00 to March 11th 00:00 is 97.5%. What I am aiming for is that the data point for March 11th in the line graph should represent the aggregated data for the past 30 days, instead of just 24 hours. This would allow us to analyze the trends in error budget total consumption trends effectively, and we also have some other reasons to need it.

As I said above, Dynatrace can provide the aggregated value in the the SLO page,

but to check the trends I need to update the timeframe on the right top corner manually. If there is an expression to provide the aggregation 30 days data, that will be great. 

[Julius_Loman]

For aggregated value as described, you need to apply the fold transformation in the formula (Required if your use metric expressions - for example to get business hours calculations properly) or use a single value visualization option such as a table or single value and just choose the fold aggregation.

Here is the second option - is this what you need? (my example is from the SaaS playground environment)

[DylanLi]

I get a strange value when using the normalized error budget metrics within a single value graph: 

 The metric expression that measures the success rate of this SLO is:

(100 *(1-calc:service.ingest.traces.errors.fivexx.rate:splitBy("dt.entity.service"):count/calc:service.ingest.traces.requestcount:splitBy("dt.entity.service"):count)):default(100,always)

I also noticed that in your demo video, there is no timeframe column as what I've seen on my end. Not sure if this configuration caused the abnormal result in the single value graph:

[DylanLi]

I am trying to use different fold transformation with the expression, but for now I haven't get the proper option there to generate the wanted graph. Maybe I will adopt Sajarra's suggestions, to export the single values out to an CSV file to create that graph.  

[PacoPorro]

Hi @DylanLi 

Did you check the metric func:slo.errorBudget.your-SLO ?

[DylanLi]

Hi @PacoPorro 
Thank you for the suggestion. I have reviewed the func:slo.errorBudget.your-SLO metric.
However, the issue I encountered is that the result is calculated using the Resolution timeframe, which defaults to 6 hours as shown in below pic. This implies that each point on the graph is based on data from the previous 6 hours. When I adjust the resolution to 30 day in the Settings to calculate based on the past 30 days, I only receive 2 points for the past month.

However, what I want to achieve is that i can get some expression, it will always calculates SLO basing on the past 30days. So that I can get a graph presents 30 points in a month and every point means the aggregated result of past 30 days. 

[Sajarra]

If you are looking something similar to above, please extract the output to a CSV format and then use that data to draw the graph..

[DylanLi]

hi Sajarra, related to the screenshot you posted, I have a question: the SLO avg listed at the left side is referring to the SLO data calculated in the past 1 day, is that correct? However, i an trying to find out a way to depict the past 30 days' SLO (SLA) based on various time points, such as viewing the error budget consumption for yesterday within the 30-day period (-31 to -1).  Is there any advices from you side about how I can get that aggregation value with an expression? 

[Sajarra]

Hi Dylan,

It's an SLO average for 30 days. I have extracted the data to a CSV format and then to excel.. Used pivot table and grouped that into a daily average and plotted the graph. Hope this answers.

[DylanLi]

Thanks @Sajarra 
This may be a solution, although I need to find a efficient way to retrieve the SLO average value for the past 30 days basing on each day