MongoDB Europe 2016 - MongoDB, Ops Manager & Docker at SNCF

MONGODB, OPS MANAGER &
DOCKER @ SNCF
Christophe TRINCAL / David TSANG-HIN-SUN / Sylvain CHAMBON
MONGODB EUROPE 2016 - NOV 15th - LONDON
FEEDBACK ON HOW TO DEPLOY MONGODB REPLICASET IN LESS
THAN 5 MINUTES ! (PROOF OF CONCEPT)

AGENDA
01.
MONGODB , PRESENTATION OF THE CONTEXT @ SNCF :
Needs & Goals
02.
INGREDIENTS / PRE-REQUISITES
03.
RECIPES & ORCHESTRATION :
how to combine Docker & Ops Manager for mongodb replicaset ?
04.
CONCLUSION:
Next steps & Improvements

ABOUT SNCF
BUSINESS PROFILE
Transport Group offering B to B and B to C services in several fields :

ABOUT SNCF
IT @ SNCF
“Production IT” is the operator serving the
IT System Departments of the group. DC #3
DC #1a
DC #1b
DC #2
Physical
servers
SmartphonesWorkstationsFixed
telephone
line
Operated
applications

NEEDS WANTS
HIGH AVAILABILITY
DELIVERY TIME
INDUSTRIALIZE
OPERATIONS
SIMPLIFICATION
SHARED SERVICES
AFFORDABLE PRICE
GOALS
RESILIENT
FASTER
CHEAPER
NEEDS AND GOALSSERVICE OFFER

TRANSLATING THE REQUIREMENTS
CHEAPER
FASTER
MORE RESILIENT
SHARED INFRASTRUCTURE
AUTOMATED DEPLOYMENT
BACKUP, ALERTING POLICY

SELECTING INGREDIENTS
SHARED
INFRASTRUCTURE
HARDWARE CLUSTER DOCKER
3 large servers (256 GB/6 TB)
3 Data Center rooms
CHEAPER
Lightweight containers
Full autonomy
Very fast startup
FASTER AUTOMATION

DOCKER
Type of container : Small/
Medium/Xlarge
Server details

SELECTING INGREDIENTS
FASTER RESILIENTAUTOMATION
BACKUP,
ALERTS
OPS MANAGER
MongoDB’s
Management
Solution

OPS MANAGER
MANAGEMENT
•  MONITORING
•  BACKUP
•  AUTOMATION
•  FULLY API-
DRIVEABLE

INGREDIENTS
1. OPS MANAGER
FASTER AND
MORE RELIABLE
DEPLOYMENTS
3 nodes for HA
2 MongoDB
Enterprise Replica
Sets
OPS 1
OPS Manager
HTTP
Service
Backup
database
(secondaire)
Application
database
(primaire)
Backup
Daemon
Head DB
Head DB
Backup
database
(primaire)
Application
database
(secondaire)
OPS 2
OPS Manager
HTTP
Service
Backup
Daemon
Head DB
Head DB
Application
database
(secondaire)
OPS 3
OPS Manager
HTTP
Service
Backup
Daemon
Head DB
Head DB
Backup
database
(secondaire)
Loadbalancer
VIP
Ops Manager

INGREDIENTS
2. HARDWARE
CLUSTER
CHEAPER
DEPLOYMENT OF
MONGODB
Linux Ubuntu 16.04
64bits /
256 GB RAM
Pool of 30 IPs pre-
allocated per server
(10Gb network interface)
Local SSD (6TB) storage
per server
And some basic orchestration : bash, python script ….

INGREDIENTS
3. DOCKER
ENGINE (V1.11)
CHEAPER, FASTER
DEPLOYMENTS
Standardised
MongoDB
deployment: 4
containers
And some basic orchestration : bash, python script ….
Mongodb replicaset

PUTTING IT ALL TOGETHER
1 2
3

Our guideline: simple, simple, simple = only 2 commands
(maybe 3 !)
l  create_replicaset.py -group sncf-test -size small –name
sncf1 -nb 3 –file passwdfile.csv -backup default -alerting
default -env prod –version 3.4 -dryrun
l  remove_replicaset.py -name replicaset-name

RECIPE & ORCHESTRATION
In the future:
upgrade_replicaset.py –name replicaset-name –size large –nb nb-
replicaset-member –env prod

create_replicaset.py –help

-group group_name
-size container’s size S/M/XL
–name replicaset-name
-nb nodes (3/3a/5/5a)
-file passwordfile.csv
-backup policy
-alerting policy
-env prod/preprod
-version 3.2.10
-dryrun (reporting only)
-help This help message

Ops Manager API – Create Group
Docker – Create Image
Capacity Planning

r = requests.post(host +
"/api/public/v1.0/groups",
auth=HTTPDigestAuth(user, key),
headers=headers,
data=json.dumps(payload))
j = r.json()

group_id = j["id"]
agent_api_key = j["agentApiKey"]
Ops Manager API –
Create Group
Capacity Planning

Ops Manager API – Create Group
Capacity Planning

Dockerfile:

FROM ubuntu-sncf:16.04
MAINTAINER ext.osmozium.david.tsang-hin-
sun@sncf.fr
RUN apt-get update
RUN apt-get install -y net-tools
RUN apt-get install -y vim-tiny
RUN apt-get install -y aptitude
RUN apt-get install -y libsasl2-2
RUN apt-get install -y ssl-cert ca-certificates
openssl
RUN apt-get install -y munin-node
ADD mongodb-mms-automation-agent-
manager_2.5.18.1647-1_amd64.deb /tmp/mongodb-mms-
automation-agent-manager_2.5.18.1647-1_amd64.deb
RUN dpkg -i /tmp/mongodb-mms-automation-agent-
manager_2.5.18.1647-1_amd64.deb
ADD automation-agent.config /etc/mongodb-mms/
automation-agent.config
RUN chown mongodb:mongodb /etc/mongodb-mms/
automation-agent.config
RUN chmod 600 /etc/mongodb-mms/automation-
agent.config
RUN mkdir /data && chown -R mongodb:mongodb /data
ADD runautomationagent.sh /runautomationagent.sh
ENTRYPOINT /runautomationagent.sh && bash

Start containers
Create and mount LVM
Volumes
Reserve IP addresses

Start containers
Create and mount LVM Volumes

Start containers
Create and mount LVM Volumes
docker run --name $replica-node-$n
--restart=always -d -h AAA.BBB.CCC.DDD -m 2G
--cpu-shares 1024 --blkio-weight=300
-p AAA.BBB.CCC.DDD:27017:27017 -ti
-v /data/replicaset1/AAA.BBB.CCC.DDD:/data:rw –v
/home/docker/vol/replicaset1/AAA.BBB.CCC.DDD:/var/
lib/mongodb-mms-automation/:rw sncf/ubuntu-
replicaset1:16.04 /bin/bash

Ops Manager Backup API
Ops Manager Alerts API
create_replicaset.py –help

-group group_name
-size container’s size S/M/XL
–name replicaset-name
-nb nodes (3/3a/5/5a)
-file passwordfile.csv
-backup policy
-alerting policy
-env prod/preprod
-version 3.2.10
-dryrun (reporting only)
-help This help message

"replicaSets": [
{
"_id": "testReplicaSet",
"members": [
{
"_id": 0,
"arbiterOnly": false,
"hidden": false,
"host": "testReplicaSet_0",
"priority": 10,
"slaveDelay": 0,
"votes": 1
},
{
"_id": 1,
"arbiterOnly": false,
"hidden": false,
"priority": 1,
"slaveDelay": 0,
"votes": 1
},
{
"_id": 2,
"arbiterOnly": true,
"hidden": false,
"priority": 1,
"slaveDelay": 0,
"votes": 1
}
]
}

Ops Manager
Automation API

Ops Manager Backup
API
Ops Manager Automation API

{
"clusterId": "CLUSTER-ID",
"dailySnapshotRetentionDays": 7,
"groupId": "BACKUP-ID",
"monthlySnapshotRetentionMonths": 13,
"pointInTimeWindowHours": 24,
"snapshotIntervalHours": 6,
"snapshotRetentionDays": 2,
"weeklySnapshotRetentionWeeks": 4
}


Ops Manager Backup API
Ops Manager Automation API

{
"eventTypeName": "MONITORING_AGENT_DOWN",
"groupId": "GROUP-ID",
"notifications": [
{
"delayMin": 0,
"emailEnabled": true,
"intervalMin": 60,
"smsEnabled": false,
"typeName": "GROUP"
}
],
"typeName": "AGENT",
}


Notify user

Scalability ?
- Add more physical hosts
Security?
- limit use of Global Owner role in Ops Manager API
Availability?
- Use of 3 Distinct DATACENTERS
Software upgrades?
- Docker 1.12 swarm enabled ? Docker API ?
- Graphical interface for true As A Service experience?
IMPROVEMENTS TO THE RECIPE

CONCLUSION
CHEAPER
FASTER
MORE RESILIENT
Several applications hosted on
shared services
End to end industrialization
Streamline architectures
Availability rate
Recovery point objective
Less than one hour data loss

NEXT STEPS & IMPROVEMENTS
Integrate to a
software
factory
Integrate into
the internal
cloud
Create
a Disaster
Recovery Plan
Next steps ….
…. and improvements

THANKS
MERCI

MongoDB Europe 2016 - MongoDB, Ops Manager & Docker at SNCF

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to MongoDB Europe 2016 - MongoDB, Ops Manager & Docker at SNCF (20)

More from MongoDB (20)

Recently uploaded (20)

MongoDB Europe 2016 - MongoDB, Ops Manager & Docker at SNCF