Keynote: Operating 600+ Mesos Servers on 7 Datacenters @Criteo
summary
Platform runs 600+ Mesos agents across seven datacenters serving ~750 production apps at up to 200K requests/sec and 40Gbps. Automation uses Chef/Consul/Apache Aurora for deployment, service discovery and scheduling; HAProxy performs L7 LB (required upgrades), logging and tracing are in place, backups and DR rehearsals exist. Networking fixes use net_cls cgroups QoS to tame noisy neighbors. Ongoing work targets CPU sets, revocability and bin packing to improve isolation and reclaim capacity after incidents like an accidental global API deletion.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 0:42Company Overview
- 1:51Transition
- 2:59Platform
- 3:14Methods
- 4:36Revision
- 4:59Automation
- 6:08Predictability
- 7:35Service Discovery
- 8:19Observability
- 10:34Networking
- 12:46Incidents
- 17:01Collaboration
- 18:13Outro
related talks
39mHow we learned to love the Data Center Operating System
18mKeynote: Mind the Gap: Bridging Cloud Native Innovation with Real-World Use
21mKeynote: CERN Experiences - Ricardo Rocha & Clenimar Filemon
19mKeynote: Building Bridges: Cloud Native and High Performance Computing
14mKeynote: The Big Leap: Conquering the Legacy System and Building a Scalable Cloud Infra...
14mKeynote: The Big Leap: Conquering the Legacy System and Building a Scalable Cloud...
25mKeynote: Infrastructure - Introduction
10mKeynote: Pushing the Limits: Scaling ArgoCD’s Repo-Server To Handle 30...