octohedron / gcp-mongodb-cluster

Launch a MongoDB cluster in Google Compute Cloud with up to thousands of shards with 2 commands.

GCP configurable MongoDB cluster

With this project you can launch your own MongoDB cluster with up to thousands of shards in Google Compute Cloud with just a few commands.

---

Features

• Easy to configure, with ansible YAML
• Set any amount of shards
• Latest MongoDB version on 4.0.X ubuntu sources list, which is4.0.12, you can change it to 4.2, or any other version in a single place in and it would set the whole cluster to that version, in gcp_common.yml
• XFS file system for the replication nodes storage
• Authentication enabled with KeyFile in all nodes
• Password protected with root user
• Configured for simple vertical scaling
• 2 Replication nodes and 1 arbiter per shard
• 1 Config server
• 1 Mongos

What do you need?

• Ansible, Python and google-auth python libraries
• A google compute cloud account with billing enabled for the API's
• A JSON service account file with the required permissions
• A compute engine project

Steps

• Edit the inventory/inventory.base.yml file, changing
  • PROJECT with your google compute engine project
  • Change the username and password
  • Set your desired zone, it's configured for frankfurt's europe-west3-b.
  • Change the ansible ssh user, which should be your configured user in Google Compute Engine so the ssh keys are copied automatically when creating the instances.

NOTE: You don't have to edit the rest of the configuration in the inventory.base.yml but it's contents are used for the inventory file that will be used, i.e. inventory.yml

• Configure the project with the configure.py script

$ python3 configure.py --shards 2 --id cl7

• Place your service account JSON file downloaded from the google compute engine website in ~/gcp_sa.json, you can change the path if you want editing the inventory.gcp.yml file

That will configure everything, first parameter of the script is the amount of shards, the second one is the cluster identifier which can't be more than 3 characters long, i.e. cl8 or any combination of 3 characters.

• Run the launch script

$ ./launch.sh

How does it work?

• The configure.py python script configures the inventory variables and the google compute inventory.
• The launch script Launches the nodes, with the ansible command ansible-playbook ./launch_nodes.yml -i inventory which
  • Launches the instances based on the ubuntu 18.04 image
  • Creates the disks, etc.

You can edit this file launch_nodes.yml if you would like a different image, type of disk, instance type, etc. For production you'd need a better instance type, i.e. not f1-micro and larger ssd disks, i.e. "pd-ssd" instead of "pd-standard", etc.

• Sets up the whole MongoDB cluster with ansible-playbook ./launch_cluster.yml -i inventory. This playbook imports all the set-up tasks in the right order, which are
  • gcp_common.yml - common provision to all nodes
  • gcp_xfs.yml - set up XFS disks
  • t_gcp_node_p.yml - provision replica set nodes
  • t_gcp_arbiter.yml - provision shard arbiters
  • t_gcp_cluster.yml - create replica sets in shards
  • t_gcp_node_auth.yml - enable auth in nodes
  • t_gcp_arb_auth.yml - enable auth in arbiters
  • t_gcp_config.yml - provision config server
  • t_gcp_mongos.yml - provision mongos server

After all of this runs, you can go to your google compute engine page, find the mongos instance external IP address, connect and check the sharding status.

For example

$ mongo --host IP_ADDRESS:27017 \
        -u 'mongouser' -p '123456' --authenticationDatabase 'admin'
MongoDB shell version v4.0.11
MongoDB server version: 4.0.12
mongos> sh.status()
  shards:
    {  "_id" : "shrs0",  "host" : "shrs0/cl8-0-ab.c.PROJECT.internal:27017,cl8-0-n.c.PROJECT.internal:27017,cl8-2-n.c.PROJECT.internal:27017",  "state" : 1 }
    {  "_id" : "shrs1",  "host" : "shrs1/cl8-1-ab.c.PROJECT.internal:27017,cl8-1-n.c.PROJECT.internal:27017,cl8-3-n.c.PROJECT.internal:27017",  "state" : 1 }
  active mongoses:
    "4.0.12" : 1
  autosplit:
    Currently enabled: yes
  balancer:
    Currently enabled:  yes
    Currently running:  no
    Failed balancer rounds in last 5 attempts:  0
    Migration Results for the last 24 hours:
      No recent migrations
  databases:
    {  "_id" : "config",  "primary" : "config",  "partitioned" : true }
      config.system.sessions
        shard key: { "_id" : 1 }

trimmed for clarirty, replace IP_ADDRESS with your mongos external IP.

Additional notes

The cluster is configured in a way in which you can reboot the nodes, after stopping all transactions of course and re-scale the instances, i.e. from f1-micro to n1-highmem-2, reboot it and enjoy the new computing power, or the other way around.

---

Something doesn't work? Open an issue.

Contributing? YES

LICENSE: MIT