From zero to hero

How to get your private cloud project to take off

Hi, I’m Maxime

  • Cloud Consultant
  • Public & Private cloud
    • OpenStack
    • Ceph
    • Kubernetes
    • DevOps

Slides

root314.com/presentations/osnd2017

QR code

Agenda

  • Hardware
  • Deployment method
  • Strategy

From zero

  • General IT knowledge
  • Heard about Cloud & OpenStack

Architecture

Architecture

Hardware

Management Switches

  • 2 x 48 x 1G RJ45 ToR: IPMI & PXE
    • Juniper EX3300-48T
    • Cisco 3048
    • Arista 7010T-48
    • EdgeCore AS4610-54T

Data Switches

  • 2 x 48 x 10G SFP+ ToR: data
    • Juniper QFX5100-48S
    • Cisco 9372PX
    • Arista 7050SX-64
    • EdgeCore AS5812-54X
  • MLAG
  • LACP on all ports

Server type Minimum Recommended
Controllers 3+1 3+1
Compute 4 8+
Block Storage 3 5
Object Storage 0 5

All servers

  • Comodity
  • 1x 1G IPMI port
  • 1x 1G PXE port
  • 2x 10G SFP+ ports
    • Intel X520
  • 2x OS SSDs 80GB+ in RAID 1
    • Intel S3520
    • Samsung SM863

Controllers

  • SuperMicro 2028TP, Dell R630, HP DL160, QCT T41S-2U
  • CPU: Dual E5-2640 v4 (10x2.4 Ghz)
  • Memory: 64GB
  • Storage:
    • 2x data SSDs RAID 1
      • Intel S3710 etc…

Compute

  • SuperMicro 2028TP, Dell R630, HP DL160, QCT T41S-2U
  • CPU: Dual E5-2683 v4 (16x2.1 Ghz)
  • Memory: 256~512GB
  • Storage: Nope

Block Storage

  • SuperMicro 1018R-WC0R, Dell R430/330, HP DL60, QCT S51G-1UL
  • CPU: Single E5-1650 v4 (6x3.6 Ghz)
  • Memory: 32GB
  • Storage: 8x 1~2TB SATA SSDs
    • Samsung SM863
    • Intel S3610 / S3520

Object Storage

  • SuperMicro 6028R-E1CR12L, Dell R730xd, HP DL180
  • CPU: Dual E5-2620 v4 (8x2.1 Ghz)
  • Memory: 64GB
  • Storage:
    • 12x 6TB HDDs - data
    • 1x 400GB NVMe SSD - journal
      • Intel P3700
      • Intel P4800X

Deployment methods

Fuel

  • Pros
    • ISO file
    • UI based install
    • Built-in PXE
  • Cons
    • Uncertain future
    • Little flexibility
    • Upgrades :(

openstack-ansible

  • Pros
    • Large user base
    • Popular tool
    • Flexible
    • Upgrades
  • Cons
    • No OS setup

Kolla

  • Pros
    • Popular tool
    • Flexible
    • Upgrades
  • Cons
    • New-ish
    • No OS setup

Operations

Monitoring: System

  • CPU
  • Disk
  • RAM
  • Network
  • Services
  • Hardware probes

Monitoring: OpenStack

  • API endpoints status
  • API error rate
  • VM capacity
  • Instances in error state
  • Neutron agents status
  • RabbitMQ cluster status
  • Galera status
  • Backup jobs

Monitoring: Ceph

  • Cluster Status
  • SMART
  • Endurance
  • Capacity per pool

Monitoring: How?

  • Existing tools
    • Nagios
    • Zabbix
    • Prometheus
  • Graphs!

Capacity Management

  • Flexibility
  • Just in time
  • Minimize unused capacity

Strategy

Don’t

  • Snowflake
  • Overpromise
  • Re-purpose VMware & SAN team

Do

  • Set realistic goals
  • General purpose
  • Start small-ish: Cloud & Team
  • Break the silo

To Hero

  • Functional Cloud (IaaS)
  • Costs similar to Public clouds
  • Flexibility

Thanks!

QR code

Credits