Treat unique label configs as separate time series in MetricRangeDataFrameSee original GitHub issue
Is your feature request related to a problem? Please describe.
MetricRangeDataFrame formats results of data type “range vector” into an easy to work with pandas dataframe. However, in this representation the time series are not “collated” or “grouped” by unique label combinations. i.e., the dataframe looks something like this:
So in order to analyze, inspect, or plot the time series corresponding to each unique label configuration, I first have to do the additional step of grouping the dataframe by labels.
Describe the solution you’d like
A dataframe where the unique label combinations are set as the MultiIndex for the dataframe, and the values are the time series values. i.e., a dataframe that looks something like this:
Describe alternatives you’ve considered
- Manually calling the
set_indexfunction with the relevant label names
- Manually doing a
groupbywith the relevant label names
This would also be more consistent with what thanos output looks like
- Created 2 years ago
- Comments:9 (6 by maintainers)
Top GitHub Comments
I’m all for this change to provide the time series grouped by unique labels as the default behavior. That said, we could use multi-indexing and maintain everything in a single DF, or we could divided the collected data into a dictionary that has a DF for each unique label set. Which may make it easier for someone to understand the data and work with the entities they are interested in. But just an alternate suggestion to consider : )
@sesheta: Closing this issue.
In response to this:
Rotten issues close after 30d of inactivity. Reopen the issue with
/reopen. Mark the issue as fresh with
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.