The execution subsystem is responsible for the execution of long running tasks on each index node. It lives in the index-service module.
It accomplishes this using the message queue and actor library, which permits program state to survive crashes and reboots.
The subsystem exposes four APIs:
- Execution API - for starting and stopping tasks, also contains miscellaneous commands
- Crawl API - for managing the crawl workflow
- Sideload API - for sideloading data
- Export API - for exporting data