DEPRECATED due to the deprecation of Neo4j HA clustering, please consider using getsocial-rnd/neo4j-aws-causal-cluster

neo4j-aws-ha-cluster

A setup for HA (High-Availability) deployment of a Neo4j Enterprise cluster on top of AWS ECS.

You can obtain Neo4j from the official website. Please contact sales@neo4j.com for Enterprise licensing.

Includes:

• Customizable CloudFormation template.
• Custom docker image on top official Neo4j image. Current version - Neo4j 3.4.6

Features

• Automatic daily backups to S3 using a slave-only instance.
• Bootstrap a cluster from a backup snapshot.
• Autoscaling (based on Memory Utilization).
• CloudWatch alerts setup.
• Bootstrap a node with an existing data volume for quick startup.
• Automatically create users+credentials for read-only and read/write access.

Prerequisites:

• Install Docker to build the image.
• AWS CLI for uploading images to ECR.

How does it work?

Neo4j graph database is deployed as a highly-available (HA) cluster with master-slave replication. 

It uses Bolt – a highly efficient, lightweight binary client-server protocol designed for database applications.

Essentially it's a Neo4j cluster with a minimum of 2 nodes (use at least 3 for HA), which is split logically into 2 ECS clusters (yet still it's 1 Neo4j cluster):

A Read-Write cluster with one master node and multiple slaves:

• Fast synchronisation between and master and nodes.
• Load Balancer keeps only a current master node in service. Hence slaves act like hot-standby in case of a failover.
• All nodes are eligible for becoming a master. Reelection will be quickly spotted by ELB.

A Read-only cluster with one slave node:

• Slower synchronisation.
• Can not become master.
• Can not accept write queries.
• Can handle complex queries without affecting performance of the R/W cluster.
• Still participates in master elections.
• Can be deployed as a different EC2 instance type for temporary heavy analytics.
• Load Balancer keeps this single node in service.
• Backups are performed on this node to avoid performance hits on R/W nodes.
• On testing environment its not present, and only a single node in the main cluster is doing the backups.

Ports open:

  - HTTP(s): 7473, 7474
  - Bolt: 7687

Usage

1. Create an ECR repository for Neo4j custom images. You will use it's ARN. (ARN looks like arn:aws:ecr:us-east-1:123456789012:repository/neo).

2. Save environment variable for use in makefile (customize them first)

   $ export NEO_ECR_REPO=<paste here ARN of your ECR repo>
   $ export NEO_AWS_REGION=<your AWS region>

3. Build Docker image and push it to your ECR:

   $ make build
   $ make push_image

4. Feel free to modify cloudformation.yml in any way you like before spinning up infrastructure, however most of the things are customizable via parameters.

5. Create a Cloud Formation stack using cloudformation.yml.

   During this step you will define all the resources you need and configure Docker image with Neo4j for ECS. Please make sure to set 2 tags for your stack (on "Options" page):

   Name: <how you name your stack>
   Environment: <your env name, e.g production>

Upgrade version

Please see detailed instructions to upgrade using this CF template.

Known Problems

• You can't restore server from a backup without a downtime. See further instructions.
• Autoscaling is hardcoded via RAM utilization (>70%). Feel free to modify for your own needs.
• Sometimes, rolling updates, that require nodes reboot, render them stuck for some time before rejoining cluster. Probaly a slower rolling update can help so that at each moment at least one node is already registered in main ELB as master.

TODO

* Parametrize autoscaling.
* Allow disabling slave-only more for simplest 1-node setups.