Version 0.1.0, Apache 2.0

Workflows that pick up where they stopped, inside your own program

FlowChassis runs your steps, writes a checkpoint before and after each one, retries what fails, undoes what can't finish and resumes every run after a crash. It is a Go library with one journal file: no server, no database, no other modules.

Quick start How it was tested
An order flow of reserve, charge and ship runs inside a program. Each step writes begin and done records to a journal file. The process is killed during charge; after the restart the engine reads the journal and carries on at charge, skipping reserve, which is already done.

0

Modules it needs besides the Go standard library

240 kills

Two recorded crash demos, 1,200 orders through 240 SIGKILLs: 899 completed, 301 compensated, none lost

Every byte

Of the journal's last record cut off in turn, and every cut opened and resumed

26,000

Steps a second with 64 workers, every checkpoint flushed to disk

Measured with FlowChassis 0.1.0 on a 2-CPU Linux cloud machine. Every number is recorded in test/results in the code repository.

Articles

about

Why the Project Is Called FlowChassis

A car’s chassis is the frame everything else bolts onto. The engine, the wheels, the body and the seats all hang from it. You don’t see it when you look at a …

Embedding FlowChassis in a product?

Teams building it into their own software, and anyone with a question about the journal, the tests or the numbers on this site, are welcome to write.

Contact