I am trying to use CircleCI’s test parallelization outlined here: https://circleci.com/docs/2.0/parallelism-faster-jobs/ but I can’t seem to get any improvements. The timing data looks like:
And I am running an rspec test step of:
- run: bundle exec parallel_test --type rspec --first-is-1 -n 2 $(circleci tests glob "spec/**/*_spec.rb" | circleci tests split --split-by=timings --timings-type=filename) # Save test results for timing analysis - store_test_results: path: test-results
I would expect over subsequent runs of the same job that the timings would have smaller and smaller amounts of variance where container 0 could run less tests and container 3 could run more, for example.