Skip to content

Lachlanmackenzie/correct run token handling

Lachlan requested to merge lachlanmackenzie/CorrectRunTokenHandling into master

Description

Some anomalies are present in the way in which both benchmarking token and results are handled in the current master during formal test runs. The principle issue is the updating (in the results cache) of the token at each benchmarking job and the automatic writing of results to the results cache. For sanity the results cached token should only be updated once the pipeline has completed (if tests fail it could suggest some corruption of the environment and hence results) and by a similar reasoning test results should only be put into the results cache once the pipeline is complete with a further safeguard to prevent erroneous runs from creating corrupt results.

Changes proposed in this merge request:

  • Use results cache token as a base for each benchmarking test and put any updated token in the pipeline cache only.
  • Write formal publishable results to pipeline cache only and then copy into the results cache when all stages have completed successfully.
  • Add debug option to test stages of "master" formal run without publishing results.
  • Validate cached run token at beginning of pipeline only (this will be used for each benchmarking job).
  • General cleanup of run token handling.

This merge request, when merged, will address issue/bug:

As per description.

Edited by Lachlan

Merge request reports