The One Algorithm That Makes Distributed Systems Stop Falling Apart When the Leader Dies
Posted by SmoothYogurtcloset65@reddit | programming | View on Reddit | 5 comments
Introductory article about Raft algorithm, which is used for leader selection in a distributed system. I see it used in tools like etcd, Consul, and Nomad for keeping configuration and service discovery consistent; in CockroachDB for storing data safely across nodes.
DoppelFrog@reddit
Also used in Kafka as Kraft.
EggCess@reddit
Which is right there in the article ...
DoppelFrog@reddit
Don't be silly. People on Reddit don't read the articles.
EggCess@reddit
Not too bad as a writeup! The writing style is a little weird maybe, and you're often missing punctuation at the end of sentences, but apart from that it's a very high-level and cursory overview of CFT consensus algorithm basics.
Note that these are non-BFT algorithms, which you kind of state by saying "when the leader node fails — can be a crash or liveness failure".
PabloZissou@reddit
Also used in NATS Jetstream