Creator of Orchestrator here. Two things I want to share and ideally get feedback on:
1) Horizontal scaling via workers
2) Flow tasks as executable plugins written in any language
Horizontal scaling via workers
The `orchestrator` binary either runs fully standalone (with the UI, scheduler, executor, etc) or as a worker that connects to a binary that runs in server mode. The latter allows horizontal scaling by simply starting the binary in worker mode on any machine.
Flow tasks as executable plugins written in any language
There's a http plugin I provide by default (others to come depending on demand). Any other task type is a separate plugin executable that provides a manifest that declares the UI elements and plugin execution code. The server (or workers) communicate with plugins over a small newline-delimited JSON protocol on stdin/stdout. Not as fast as in-process, but extensible without any code modifications in Orchestrator.
Creator of Orchestrator here. Two things I want to share and ideally get feedback on:
1) Horizontal scaling via workers 2) Flow tasks as executable plugins written in any language
Horizontal scaling via workers The `orchestrator` binary either runs fully standalone (with the UI, scheduler, executor, etc) or as a worker that connects to a binary that runs in server mode. The latter allows horizontal scaling by simply starting the binary in worker mode on any machine.
Flow tasks as executable plugins written in any language There's a http plugin I provide by default (others to come depending on demand). Any other task type is a separate plugin executable that provides a manifest that declares the UI elements and plugin execution code. The server (or workers) communicate with plugins over a small newline-delimited JSON protocol on stdin/stdout. Not as fast as in-process, but extensible without any code modifications in Orchestrator.