Gareth Ellis

Investigate the time taken, and number of benchmarks in core

6

This is the starting place to begin running core benchmarks on a regular basis. First steps: Put together a benchmark run that summarises time to run each suite as well...

Community benchmark run.sh changes

1

I'm putting this here to cover a few changes I think are necessary to the community benchmark job in jenkins. Currently the version we run is from the core-benchmarks branch...

Status of running node core benchmarks

29

Since the PR was closed once I'd merged in my changes, I figured I'd open this to track current progress on getting this working. This is a continuation of #58...

Update create-track in rally

3

Add option to increase batch size when creating a track to speed up download of data. Update track layout to adhere to best practices Creating tracks from large corpus can...

Bulk http status & 429 retries

1

Future direction: - need to avoid messy nested fields - service_time should be a sum of all retries - follow approach of other products, such as logstash

Cast batch size as int

1

Spotted this whilst making some other changes to create-track.

Initial support for proxies. The aim is to rely on existing support from requests or aiohttp. Next steps: - more thorough testing - documentation

We currently cant know if a track is in test mode within custom runner code. This adds the information for runners - I am still working on for ParamSources.

Originally a decision was made that rally should always update existing index templates - the reasoning was that the fields in the index could well be updated regularly. We are...

When trying to push Elasticsearch to extremes, some requests may not succeed due to e.g circuit breakers triggering 429s amongst other things. To ensure that we ingest _all_ documents that...