data-visualizer script graph labels overlapping

[Deblintrake09]

Target version	Related issue	Related PR/dev branch
-	-	4.8.0

Description

During research for https://github.com/wazuh/wazuh/issues/20781 it was detected that the script outputs graphs with overlapping labels. This was replicated on Ubuntu 22.04 and Windows 11 instances.

Module	Windows	Ubuntu
python	3.12.1	3.10.12
matplotlib	3.8.2	3.8.2

Research is needed on cause and apply fixes as needed.

[MARCOSD4]

Update

I have read and reviewed the issue and all the other issues it depends on, related to the workload, to better understand the problem. Also, I have reviewed the data-visualizer script.

[MARCOSD4]

Update

I have reviewed how plots are created and performed various tests in my local environment using previously created artifacts but have not been able to replicate the problem.

I have launched this build to test with new artifacts: https://ci.wazuh.info/job/CLUSTER-Workload_benchmarks_metrics/397/

[MARCOSD4]

Update

I have launched several builds to test the generated artifacts and performed several tests but have not been able to replicate the problem because the time range of the tests is too small. I have launched another build to get data with a larger time range and it should be tested when finished: https://ci.wazuh.info/job/CLUSTER-Workload_benchmarks_metrics/403/.

[MARCOSD4]

Update

In the initial configuration of the visualization.py script, it was set by default that the range of values of the x-axis labels had an interval of 1 minute. This works correctly when the time range of the tests is small, but if we increase it, to approximately close to 6 hours, the labels overlap and we get graphs like the one shown in this issue.

After trying different alternatives, it has been decided to leave a fixed number of labels (30), whose intervals will be automatically adjusted according to the time of the question test, thus avoiding the labels overlap whatever the duration of the tests. This way we will always have 30 labels, which are enough for the graph to be representative, and they will never overlap with each other.

Different tests have been carried out with 18 minutes, 2 hours, and 6 hours tests and the results are as follows:

6 hours


2 hours


18 minutes


Finally, a build has been launched to test that everything works correctly: https://ci.wazuh.info/job/CLUSTER-Workload_benchmarks_metrics/415/

Artifacts: Artifacts.tgz


[rafabailon]

LGTM

[davidjiglesias]

LGTM