opensearch-project/opensearch-benchmark-workloads

[BUG] Unused search operations in the nyc_taxis workload

Aperta

#67 aperta il 21 mar 2023

 (1 commento) (0 reazioni) (1 assegnatario)Python (137 fork)auto 404
Low Prioritybuggood first issue

Metriche repository

Star
 (33 stelle)
Metriche merge PR
 (Merge medio 5g 19h) (2 PR mergiate in 30 g)

Descrizione

What is the bug?

The 'nyc_taxis' workload contains the following search operations (defined here).

  1. default
  2. range
  3. distance_amount_agg
  4. autohisto_agg
  5. date_histogram_agg
  6. date_histogram_calendar_interval
  7. date_histogram_calendar_interval_with_tz
  8. date_histogram_fixed_interval
  9. date_histogram_fixed_interval_with_tz
  10. date_histogram_fixed_interval_with_metrics
  11. auto_date_histogram
  12. auto_date_histogram_with_tz
  13. auto_date_histogram_with_metrics

Out of these, only 1–5 are being used in a test-procedure (append-no-conflicts) while they remaining are not mapped to any test-procedure (defined here). This reduces our coverage for 'aggregation' queries.

It appears that the 'date-histogram' test-procedure that executed these unmapped operations was incorrectly removed in https://github.com/opensearch-project/opensearch-benchmark-workloads/commit/cba4e45dda37ac03abbd3c9dd4532475dac355e9, thinking it was part of X-pack code, which it wasn't.

What is the expected behavior?

The missing test-procedure should be added back. I have confirmed that these operations execute successfully on open-source OpenSearch.

What is your host/environment?

Ubuntu 22.04.2 LTS (GNU/Linux 5.15.0-1031-aws x86_64) opensearch-benchmark 0.2.0

Guida contributor