Unlocking Diego: Achieving Distributed Robustness Through Simplicity and Reliability
summary
Cloud Foundry’s Consul + RAFT setup proves fragile under repeated BOSH redeploys, causing leader election failures and split brains that drop routes. Diego replaces Consul locks with Locket on relational DB to avoid RAFT split‑brain, moves from a single global router emitter to local route-emitter per cell so route loss affects only a cell, and is prototyping BOSH DNS to replace Consul DNS. These changes are deployed and reduce Consul-caused outages and orchestration pain.
AI-written summary · about
chapters (11)
- 0:00Intro
- 0:47Background • Diego is the runtime / scheduler for Cloud Foundry • Diego is a distributed system in nature, hence its components
- 1:57What is Consul?
- 2:59Consul and Cloud Foundry 1. Service Discovery (uses Consul ONS)
- 3:57Problem Statement Why is Consul an issue for Cloud Foundry?
- 4:42Consul and the RAFT algorithm
- 7:36How are BOSH and Consul in conflict? 1. Bosh roling deploys takes a leader down
- 9:09Consul and Diego
- 11:48Consul failure and Diego
- 13:22BBS and Auctioneer
- 19:17Wrapping up 1. Consul is good but not combined with BOSH
related talks
47mDistributed Systems: When Democracy and Computer Science Come Together
46mWhat can go wrong in a distributed system - experience from...
33mWhy Are Distributed Systems So Hard?
30mDiego Runtime Overview
43mBuild Your Own Distributed System Using Go
45mDo Distributed Systems they said. It'll be fun they said.
20mTaking Cloud to Extremes: Scaled-Down, Highly-Available, and Mission-Critical
43mCreating Fault Tolerant Services on Mesos