absearch

The AB Search service provides 5 commands:

• absearch-server: runs the service
• absearch-upload: uploads the config located in data in S3
• absearch-check: validates the config using the schema
• absearch-redis-dump: dumps the Redis counters
• absearch-redis-load: loads the counters into Redis

Overview

• config.json contains the search settings per locale and territory, and also cohorts for a/b testing
• Editors change the config JSON file in Github
• After some sanity check, a script uploads that file in S3
• Firefox calls the service to get the search settings, providing a locale & territory
• The app grabs them from S3 (with a local memory copy udpated every hour) and distribute users among cohorts
• The app uses Redis to keep track of cohort counters, that are used for the distribution

How to import/export in Redis

Use absearch-redis-dump to export in a file, the script takes the config file as an argument:

$ absearch-redis-dump absearch.ini > redis-dump

The produced file contains all the counters extracted from Redis. To reinject them into Redis, use absearch-redis-load. It takes the config file and the data file as params:

$ absearch-redis-load absearch.ini redis-dump
Loading 'redis-dump' into Redis
Done

Metrics

The following metrics are produced by the app in statsd:

Name	Type	Description
absearch.redis.incr	timer	on every redis increment
absearch.redis.get	timer	on every redis lookup
absearch.s3.set_file	timer	file uploaded in S3
absearch.s3.get_file	timer	file downloaded from S3
absearch.add_user_to_cohort	timer	on every GET w/ a new cohort set
absearch.enrolled	counter	on every cohort enrollment (*)
absearch.discarded	counter	on every cohort discard () (*)
absearch.refreshed	counter	on every call to get cohort settings (*)
absearch.get_cohort_settings	timer	on every GET to get back cohort settings

() if you use Datadog, the tag will contain the cohort name, prefixed by the locale and territory. (*) unknown cohorts will increment this counter as well