From Mesos to Kubernetes
summary
Migration from Mesos after three years moves a multi-cluster fleet (eight clusters) onto Kubernetes while keeping an adapter layer so Mesos and K8s engines coexist and APIs translate. Container networking avoids a single IP allocator by using a host-gateway CNI with preset routes, cutting the earlier 16% network difference down to zero. Storage uses per-container LVM volumes to guarantee isolated size and I/O. Monitoring and logs run with a Falcon push-agent and delayed-log flush; StackStorm automates root-cause workflows and self-heal actions.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 1:48Container Network
- 5:18Storage Architecture
- 10:12Mesos debug mode
- 13:45Mesos debug log
- 15:27Secure private login
- 16:39Dynamics of security
- 18:42Monitoring
- 19:35Self discovery
- 20:17Monitoring system
- 23:02Cutting
- 24:52Container Engine
- 25:18TCOS OS
- 27:40Service Mash
- 29:22Chaos Simulation
- 31:21DCOs
- 32:22Decision Tree
- 34:00SelfCuring
- 35:36Outro
related talks
31mMoving from Mesos to Kubernetes Without Anyone Noticing
24mMigration of NCSOFT Game Servers to Kubernetes
31mMigrating Legacy Infrastructure to Kubernetes: Olark’s Lessons From the Trenches
36mOnly Slightly Bent: Uber’s Kubernetes Migration Journey for Microservices
35mClusters as Cattle: How to Seamlessly Migrate Apps Across Kubernetes Clusters
01h 25mTutorial: Migration 101: From VMs to Kubernetes
20mKeynote: The Challenges of Migrating 150+ Microservices to Kubernetes
29mFrom EC2 to Kubernetes Takeaways from a Brownfield Migration