Deploy mongodb replicaset servers with Docker on different physical servers - mongodb

Question

I'm trying to deploy a mongodb replicaset using docker.
I managed to do it on a same server by executing this :
docker run -d --expose 27017 --name mongodbmycompany1 dockerfile/mongodb mongod --replSet rsmycompa
docker run -d --expose 27017 --name mongodbmycompany2 dockerfile/mongodb mongod --replSet rsacommeassure
docker run -d --expose 27017 --name mongodbmycompany3 dockerfile/mongodb mongod --replSet rsacommeassure
MONGODB1=$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' mongodbmycompany1)
MONGODB2=$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' mongodbmycompany2)
MONGODB3=$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' mongodbmycompany3)
echo $MONGODB1
echo $MONGODB2
echo $MONGODB3
echo "Mongodb Replicaset init"
docker exec mongodbmycompany1 mongo 127.0.0.1:27017/mycompany --eval 'if(!rs.conf()) { rs.initiate(); cfg = rs.conf(); cfg.members[0].host = "'$MONGODB1':27017"; rs.reconfig(cfg); rs.add("'$MONGODB2':27017"); rs.add("'$MONGODB3':27017"); } rs.status();'
It's working as expected. My replicaset is initialized and my mongodb resultset config contains my 3 servers identified by their internal IP address.
It's not perfect as I'd prefer to use servers names but I didn't manage to do it.
Docker only populate each /etc/hosts file with servers names passed at image launch with --link parameter. If i add a new server while others are running. Those servers won't ping the new server.
Now I have another question. In production, having a lot of Mongodb docker image running on a same physical server is possible but it's not safe :
- if my physical server falls down, i lose all my Mongodb replicas and my service is down
- if my physical server uses internal storage, all my docker images use the same disk... and I'm going to have IO problems.
So my question is : How can I deploy many mongodb replicas on multiple physical servers ? How those mongodb replicas can communicate with each others (primary and secondaries servers can change) while they are on different servers or even on different datacenters ?

Answer 1

Let's assume:
you have 3 different docker hosts (servers), with IPs 10.1.1.101, 10.1.1.102, 10.1.1.103
want to deploy a single replica set called rsacommeassure
Dockerfiles for mongodb expose port 27017
all servers are in a trusted zone and can talk to each other
First let's start mongodb containers on each server (10.1.1.101 ~$ is used for command prompt):
10.1.1.101 ~$ docker run -d -p 27017:27017 --name mongodbmycompany1 dockerfile/mongodb mongod --replSet rsacommeassure
10.1.1.102 ~$ docker run -d -p 27017:27017 --name mongodbmycompany2 dockerfile/mongodb mongod --replSet rsacommeassure
10.1.1.103 ~$ docker run -d -p 27017:27017 --name mongodbmycompany3 dockerfile/mongodb mongod --replSet rsacommeassure
-p 27017:27017 exposes port 27017 on the host IP so mongo is accessible on servers' host IP address.
Then you need to initiate the replica set, so just run this against a mongodb container (I'll pick server1 here):
your_laptop ~$ > mongo --host 10.1.1.101
MongoDB shell version: 2.6.9
connecting to: test
> rs.initiate()
> cfg = rs.conf()
> cfg.members[0].host = "10.1.1.101:27017"
> rs.reconfig(cfg)
> rs.add("10.1.1.102:27017")
> rs.add("10.1.1.103:27017")
> rs.status();
The IPs are local but it works with global as well as long as the servers can talk to each other (VPN, firewall, DMZ, whatever). Btw you should consider security carefully.

Answer 2

I've created a Replica Set on different physical servers using docker-machine and virtual box as a driver:
$ docker-machine create --driver virtualbox server1
$ docker-machine create --driver virtualbox server2
$ docker-machine create --driver virtualbox server3
Open 3 different terminals, in each
$(Terminal1) eval "$(docker-machine env server1)"
$(Terminal2) eval "$(docker-machine env server2)"
$(Terminal3) eval "$(docker-machine env server3)"
In each terminal:
$(Terminal1) docker run -d -p 27017:27017 --name mongoClient1 mongo mongod --replSet r1
$(Terminal2) docker run -d -p 27017:27017 --name mongoClient2 mongo mongod --replSet r1
$(Terminal3) docker run -d -p 27017:27017 --name mongoClient3 mongo mongod --replSet r1
Go in VirtualBox -> on each environment(server1,server2,server3) -> Setting -> Network -> Adapter 1 -> Port Forwarding. Create a new rule Protocol TCP, Host Port - 27017, Guest Port - 27017, leave Host Ip and Guest Ip empty
Now restart all the environments, you can do this from the VirtualBox or from the terminal, from terminal just run:
$(Terminal1) docker-machine restart server1
$(Terminal2) docker-machine restart server2
$(Terminal3) docker-machine restart server3
Restart the containers:
$(Terminal1) docker start mongoClient1
$(Terminal2) docker start mongoClient2
$(Terminal3) docker start mongoClient3
Now the containers should be running, you can check them by running
$ docker ps in each terminal
Get into the first container's(or another) Mongo Shell
$(Terminal1) docker exec -it mongoClient1 mongo
// now we are in the Mongo Shell
$(Mongo Shell) rs.initiate()
$(Mongo Shell) cfg = rs.conf()
$(Mongo Shell) cfg.members[0].host = <server1's Ip Address>
// you should get server1's Ip Address by running $ docker-machine ls, mine was 192.168.99.100
$(Mongo Shell) rs.reconfig(cfg)
$(Mongo Shell Primary) rs.add("<server2's Ip Address>:27017")
// now we added a Secondary
$(Mongo Shell Primary) rs.add("<server3's Ip Address>:27017", true)
// now we added an Arbiter
$(Mongo Shell Primary) use planes
// now we create a new database
$(Mongo Shell Primary) db.tranporters.insert({name:'Boeing'})
// create a new collection
$(Mongo Shell Primary) db.tranporters.find()
// we obtain the inserted plane
To connect to a Secondary, you can either:
$(Terminal2) docker exec -it mongoClient2 mongo planes
// or
$(Mongo Shell Primary) db = connect ("<server2's Ip Address>:27017/planes")
Now we are in the Mongo Shell of a Secondary
$(Mongo Shell Secondary) rs.slaveOk()
// to allow readings from the Shell
$(Mongo Shell Secondary)db.tranporters.find()
// should return inserted plane

Answer 3

You could use "Weave - the Docker network" to resolve your problem easily.
Weave creates an overlay network that joins containers on different hosts, even at different cloud providers. Weave also supplies a DNS service that lets you find containers by name within the Weave network.

Answer 4

#Here stop already running instances
docker stop m1 m2 m3
#Cleanup of the volumes
docker rm -f m1 m2 m3
# Start MongoDB services optimised for faster startup
docker run -dP --name m1 mongo mongod --replSet "r1" --noprealloc --smallfiles --nojournal --syncdelay 0
docker run -dP --name m2 mongo mongod --replSet "r1" --noprealloc --smallfiles --nojournal --syncdelay 0
docker run -dP --name m3 mongo mongod --replSet "r1" --noprealloc --smallfiles --nojournal --syncdelay 0
export M1_ADDRESS=`docker inspect --format '{{ .NetworkSettings.IPAddress }}' m1`
export M2_ADDRESS=`docker inspect --format '{{ .NetworkSettings.IPAddress }}' m2`
export M3_ADDRESS=`docker inspect --format '{{ .NetworkSettings.IPAddress }}' m3`
docker exec m1 mongo --eval "rs.initiate();"
docker exec m1 mongo --eval "cfg = rs.conf(); cfg.members[0].host = '$M1_ADDRESS:27017'; rs.reconfig(cfg);"
docker exec m1 mongo --eval "rs.add('$M2_ADDRESS:27017');rs.add('$M3_ADDRESS:27017')"
# Check if everything is fine.
docker exec m1 mongo --eval "rs.status();"

Answer 5

I scripted a docker-image that sets up a mongodb replicaSet for any number of containers and automates scaling as you add more containers. check out the github github repo or docker registry
Use Docker Compose
setup your docker-compose.yml
version: "2"
services:
<your_service_name>:
image: rollymaduk/mongo-replica:local
environment:
REPLICA_NAME: "<your_replica_name>"
volumes:
- /var/config:/var/config
run in command line
docker-compose up
scale up to more containers
docker-compose scale <your_service_name>=3
Docker Cloud
To deploy mongo-db replicaSet using docker-cloud, set up a stack
file
stack.yml: stack file not requiring shared volumes
<service_name>:
image: rollymaduk/mongo-replica:cloud
deployment_strategy: high_availability
target_num_containers: 3
environment:
REPLICA_NAME: <your_replica_name>
DOCKERCLOUD_AUTH: <your_docker_auth_key>
stack.yml: stack file requiring shared volume
<service_name>:
image: rollymaduk/mongo-replica:local
deployment_strategy: high_availability
target_num_containers: 3
volumes:
- /var/config:/var/config
environment:
REPLICA_NAME: <your_replica_name>
using docker-cloud cli run in command line (If stack file does not exist in cloud)
docker-cloud stack create --name <your_stack_name> -f <your_stack_file>
docker-cloud stack start <your_stack_name>
using docker-cloud cli run in command line (If stack file already exists in cloud)
docker-cloud stack update -f <your_stack_file> <your_stack_name>
docker-cloud stack redeploy <your_stack_name>
Important: you must specify a shared volume and mount it to the container's
config directory [default: /var/config ] the example below mounts a host
directory /var/config to the container's config volume.
You can use any of the docker recommended ways of sharing volumes between
containers i.e mounting a host directory or data volume container),just
you specify the correct path for the config volume (/var/config).
_to change default config directory use the environment variable
--CONFIG_DIR to setup your container. Make sure to update host volume
path to reflect your new --CONFIG_DIR name_