akst / kira Goto Github PK

Currently everything is massive files, let try and change that.

Time out on lack of message passing makes no sense from the users perspective, letting the user provide a timeout on a task by task basis makes a lot more sense, as you be prepared to wait a lot longer for some tasks, and some tasks may rely on external service whose reliability is not something you can depend on.

Continuing on from #2, we'll want to document more advanced use cases.

• Tear down during rollback
• Reattmpting after task failure
• Reattmpting in rollback

Lets make it happen.

Currently the documentation is mostly non existent, and it's not obvious what functions are intended to do what.

The retry logic should be taking place in its own process, and it should be asynchronous & not block the state machine from proceeding.

We'll want examples of trivial and non trivial usage.