elastic / rally

Macrobenchmarking framework for Elasticsearch
Apache License 2.0
1.95k stars 313 forks source link

Simplify test mode with small document sets #1792

Closed gbanasiak closed 8 months ago

gbanasiak commented 11 months ago

https://github.com/elastic/rally-tracks/pull/469 adds very small indices for enrichment purposes. This scenario is not supported by test mode which requires -1k files with 1000 documents. This leads to an awkward workaround where original corpus gets dummy content just to reach 1000 documents in -1k files.

This PR addresses this limitation by skipping the -1k requirement for document sets with 1000 and less documents. In such case, original document set file is taken when running in test mode.

pquentin commented 11 months ago

Is this closing https://github.com/elastic/rally/issues/1354?

gbanasiak commented 11 months ago

Is this closing https://github.com/elastic/rally/issues/1354?

Yes although addressing it wasn't the initial intention here - I wasn't aware this issue exists. Looking at the discussion there the novelty now is the arrival of a case where we have both big/regular corpus ("main" index) and very small corpus (enrichment indices) where test mode is still useful. This scenario wasn't thought of back in 2021, and I think it tips the scale towards the change of test mode behaviour.