After merging #57241 we see the following spike in aggregation latency in the nightly charts (starting 20200528):
The list of commits between 20200527 -> 20200528 is: 16be0e6...9f6bc68
and based on analysis, it can be attributed to #57241.
The specific regression is on the country_agg-uncached query on geonames and it reproduces regardless of license type.
The following Rally invocations can be used to identify the regression:
Before #57241 i.e. 6f79750
esrally --track=geonames --challenge=append-no-conflicts --include-tasks="delete-index,create-index,check-cluster-health,index-append,refresh-after-index,force-merge,refresh-after-force-merge,wait-until-merges-finish,index-stats,node-stats,country_agg_uncached" --runtime-jdk=bundled --car=defaults --revision=6f79750 --user-tag="commit:before_make_global_ords_terms_simpler_to_understand" --on-error=abort
Using suspect commit: 974d236
esrally --track=geonames --challenge=append-no-conflicts --include-tasks="delete-index,create-index,check-cluster-health,index-append,refresh-after-index,force-merge,refresh-after-force-merge,wait-until-merges-finish,index-stats,node-stats,country_agg_uncached" --runtime-jdk=bundled --car=defaults --revision=974d236 --user-tag="commit:make_global_ords_terms_simpler_to_understand" --on-error=abort
After merging #57241 we see the following spike in aggregation latency in the nightly charts (starting 20200528):
The list of commits between 20200527 -> 20200528 is: 16be0e6...9f6bc68
and based on analysis, it can be attributed to #57241.
The specific regression is on the country_agg-uncached query on geonames and it reproduces regardless of license type.
The following Rally invocations can be used to identify the regression:
Before #57241 i.e. 6f79750
Using suspect commit: 974d236