odyssey VS pgcat

This is /extremely/ uncharitable. It depends more on the project.

For example, both Go and .NET are very well written with comprehensible documentation. Many projects have also been absorbed by the Apache Foundation or other FOSS initiatives.

Every FANG has boat loads of teams working on umpteen different open source packages. The quality naturally varies.

Let me pick two more esoteric projects as a point of comparison.

Yandex Odyssey (https://github.com/yandex/odyssey) an advanced multi-threaded PostgreSQL connection pooler and request router. Figuring out how exactly and when to use this is not quite clear. There is no "getting started" guide for this package. There is barely any explanation for how it works or what it does.

pg_auto_failover (https://github.com/citusdata/pg_auto_failover) run by Citus (owned by Microsoft) monitors and manages automated failover for a Postgres cluster. This repo even has diagrams explaining the workflow and complete instructions.

Odyssey, developed by Yandex, is a newer solution for pooling. While Pgbouncer is a single-threaded pooler, Odyssey supports multi-core and multi-threaded processing. The tool has the following features:

You should check out https://github.com/levkk/pgcat

I created a pool called "yugabyte" (pools.yugabyte), with pool mode 'session', so that a session remains connected to a backend. I created a pool of servers in pools.yugabyte.shards.0 (yugabyte shards pool 0), with all three of the YSQL nodes in the YugabyteDB cluster, and gave one the role of "primary" and the two others the role of "replica". For YugabyteDB YSQL this does not make sense. Currently, pgcat allows to have the roles to be all "replica", but not have more than one "primary".

pgcat is pgbouncer rewritten in Rust with support for load balancing between replicas, failover in case a replica fails a health check, and sharding at the pooler level.

Show HN: PgCat, Postgres pooler with sharding, load balancing and failover\ (38 comments)

This is actually pretty amazing. I wouldn't normally trust a single author implemented project to sit between my application and my database, since that is a rather critical single point of failure. However, a grep that fails to find `unsafe` in a rust code base lets me know that this code is in fact pretty safe[1]. I don't need to trust the author, as much as I trust the Rust core team, and LLVM.

Since I can see basic functionality exercised in the test suites, and trust that Rust has validated its more general correctness, this gives me more confidence in your project than I would historically expect to be able to gather in such a short amount of time.

1: https://github.com/levkk/pgcat/search?q=unsafe