Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

0

Share

Download to read offline

Scaling docker @ovh

Download to read offline

How ovh went from 20 to 2500 containers in production

Related Books

Free with a 30 day trial from Scribd

See all

Related Audiobooks

Free with a 30 day trial from Scribd

See all
  • Be the first to like this

Scaling docker @ovh

  1. 1. Scaling docker @OVH Timothée GERMAIN devops@OVH From 20 to 2500 running containers
  2. 2. About me Timothée GERMAIN Devops for Containers team @ OVH since 2015 Work with containers since 2014 Twitter : DarkNihilius1
  3. 3. Docker @ OVH ● 8 clusters ○ Internal production : 5 ○ External beta/gamma : 2 ○ Preproduction : 1 ● 2500 running containers ● 800 Hosts ● 3000 cpu cores ● 12TB RAM ● 2 docker registry (internal & external) ● 100s of deployment per day ● 5 people
  4. 4. The beginning : 1 - 20 containers Everything fit on 1 host + Easy to deploy docker run + Easy to build docker build + Easy to debug docker exec/logs/stats - No high availability - Not scalable
  5. 5. Scaling up : +20 containers, 1 team You need an orchestrator : ● Docker swarm (Docker inc.) ● Mesos & Marathon (Apache Software Foundation) ● Kubernetes (Google then CNCF) Don’t fit on 1 host ? Need high availability ?
  6. 6. Mesos & marathon in 60 seconds Linux kernel : provide abstraction of hardware Linux kernel Your application Linux scheduler Scheduler : say which application run when with which resources
  7. 7. Mesos & marathon in 60 seconds Mesos : provide abstraction of multiple hosts Mesos Your (Docker) application Framework Framework (Marathon) : say which application run when with which resources
  8. 8. Needs : ● Isolate team/client ● Be cost efficient, mutualize what we can Scaling up : 20+ containers, 2+ team More team !
  9. 9. Scaling up : 20+ containers, 2+ team Marathon : no multitenancy out of the box The solutions we try : 1) One marathon per team 2) One marathon with logical isolation (middleware) 3) One mesos & marathon per team
  10. 10. Multitenancy : One marathon per team + Good isolation + Mesos mutualization + Fair blast radius - Does not scale well Mesos Marathon B Marathon CMarathon A
  11. 11. Multitenancy : One marathon and logical isolation + Scale + Give us control on user input - Less isolation - Lots of maintenance - Hard to keep up to date with marathon API - Big blast radius Mesos Marathon Middleware
  12. 12. Multitenancy : One marathon & mesos cluster per team + Scale even better + Little blast radius + Best isolation + Modular - Less/no mutualisation - Costly - Lot of parts to deploy and maintain Mesos B Marathon B Mesos A Marathon A
  13. 13. Deploy mesos and marathon using mesos and marathon Lot of parts to deploy and maintain ? we already have a tool for that : Mesos and Marathon ! Mesos B Marathon B Mesos A Marathon A Mesos Admin Marathon Admin
  14. 14. Questions ? Thanks for your time
  15. 15. Log and metrics : 1 - 20 containers Docker stats Docker logs Direct calls on docker daemon
  16. 16. Log and metrics : +20 containers, 2+ team Not docker logs nor docker stats

How ovh went from 20 to 2500 containers in production

Views

Total views

655

On Slideshare

0

From embeds

0

Number of embeds

1

Actions

Downloads

2

Shares

0

Comments

0

Likes

0

×