# Multi-Node Deployment

This documentation describes how to run loom in a multi-node setup.

# Installation

The following steps need to be executed on each node:

1. Choose a working directory of your choice. In this example, we are using /home/ubuntu

   cd /home/ubuntu
   

2. Download the binaries:

   curl https://raw.githubusercontent.com/loomnetwork/loom-sdk-documentation/master/scripts/get_loom.sh | sh
   

3. Execute ./loom init in the working directory to initialize config files.

4. Add loom.yml in the working directory:

   DPOSVersion: 3
   

# Configuration

There are two genesis.json files that need to be combined.

# genesis.json #1 - in the working directory

The genesis.json that was generated should look something like below. You should NOT copy the file below as it is an example. Use the file generated in your working directory.

{
  "contracts": [
    {
      "vm": "plugin",
      "format": "plugin",
      "name": "coin",
      "location": "coin:1.0.0",
      "init": null
    },
    {
      "vm": "plugin",
      "format": "plugin",
      "name": "dposV3",
      "location": "dposV3:3.0.0",
      "init": {
        "params": {
          "validatorCount": "21"
        },
        "validators": [
          {
            "pubKey": "2MysikRZ8Yzk3KPDVEl/g2tHSyX0i3DGrAMwtDcYH10=",
            "power": "10"
          }
        ]
      }
    },
    {
      "vm": "plugin",
      "format": "plugin",
      "name": "addressmapper",
      "location": "addressmapper:0.1.0",
      "init": null
    },
    {
      "vm": "plugin",
      "format": "plugin",
      "name": "chainconfig",
      "location": "chainconfig:1.0.0",
      "init": {
        "owner": {
          "chainId": "default",
          "local": "aMt0mxDIxz5MCYKp9c0jEzG1en8="
        },
        "params": {
          "voteThreshold": "67",
          "numBlockConfirmations": "10"
        },
        "features": [
          {
            "name": "test",
            "status": "WAITING"
          }
        ]
      }
    }
  ]
}

Next, collect all validators from each node, combine them into an array, and save everything to a new file. This old file will now need to be replaced with the new combined file. Do this for all nodes. For a two-node cluster, the validators array should look something like this:

  "validators": [
    {
      "pubKey": "2MysikRZ8Yzk3KPDVEl/g2tHSyX0i3DGrAMwtDcYH10=",
      "power": "10"
    },
    {
      "pubKey": "gCn5WayR3cgQjNlNXYiBSYgQ3c1pGIsFWajVGczByZulGa09mb",
      "power": "10"
    }
  ]

# genesis.json #2 - inside chaindata/config

You will find it at chaindata/config/genesis.json. It is not to be confused with the one in the working directory. You should NOT copy the file below as it is an example and it should look something like:

{
  "genesis_time": "2019-06-20T08:58:17.011337021Z",
  "chain_id": "default",
  "consensus_params": {
    "block_size": {
      "max_bytes": "22020096",
      "max_gas": "-1"
    },
    "evidence": {
      "max_age": "100000"
    },
    "validator": {
      "pub_key_types": [
        "ed25519"
      ]
    }
  },
  "validators": [
    {
      "address": "825F1AE812A4395EFF0F88A032AAB2CE42F120EE",
      "pub_key": {
        "type": "tendermint/PubKeyEd25519",
        "value": "2MysikRZ8Yzk3KPDVEl/g2tHSyX0i3DGrAMwtDcYH10="
      },
      "power": "10",
      "name": ""
    }
  ],
  "app_hash": ""
}

Next, collect all the validators from each node, combine them into an array, and save everything to a new file. The old file will now need to be replaced with the new combined file. Do this for all nodes. For a two-node cluster, the validators array should look something like this:

  "validators": [
    {
      "address": "825F1AE812A4395EFF0F88A032AAB2CE42F120EE",
      "pub_key": {
        "type": "tendermint/PubKeyEd25519",
        "value": "2MysikRZ8Yzk3KPDVEl/g2tHSyX0i3DGrAMwtDcYH10="
      },
      "power": "10",
      "name": ""
    },
    {
      "address": "825F1AE812A4395EFF0F88A032AAB2CE42F120EE",
      "pub_key": {
        "type": "tendermint/PubKeyEd25519",
        "value": "gCn5WayR3cgQjNlNXYiBSYgQ3c1pGIsFWajVGczByZulGa09mb"
      },
      "power": "10",
      "name": ""
    }
  ],

# Running

First, we need to get node keys from each node. Go to the working directory and run loom nodekey:

loom nodekey

You should see something like this:

47cd3e4cc27ac621ff8bc59b776fa228adab827e

Do remember to clearly make a note which node key is for which node. Also important is the private IP (or any IP that are available for the nodes to communicate with each other). Generally, in a cloud environment, we use public IPs for security and latency reasons.

Now, let's use an example with 4 nodes:

To run loom, we need to tell each node about its peers. The general format is:

loom run --persistent-peers tcp://<node1_key>@<node1_ip>:26656,tcp://<node2_key>@<node2_ip>:26656,...tcp://<nodeN_key>@<nodeN_ip>:26656

Let's see examples by using the table above.

On node 1:

loom run --persistent-peers tcp://[email protected]:26656,tcp://[email protected]:26656,tcp://[email protected]:26656

On node 2:

loom run --persistent-peers tcp://[email protected]:26656,tcp://[email protected]:26656,tcp://[email protected]:26656

The same goes for node 3 and node 4. We exclude the node's own key and IP address.

Please remember that all commands need to be executed from within the working directory.

# systemd Startup Script

The following startup script can be used to control the service using systemd. Make changes to WorkingDirectory and/or ExecStart to reflect your setup.

Notice ExecStart, it is constructed using the same concept from the previous section when running loom directly. This means each node has a different startup script.

[Unit]
Description=Loom
After=network.target

[Service]
Type=simple
User=ubuntu
WorkingDirectory=/home/ubuntu
ExecStart=/home/ubuntu/loom run --persistent-peers tcp://[email protected]:26656,tcp://[email protected]:26656,tcp://[email protected]:26656
Restart=always
RestartSec=2
StartLimitInterval=0
LimitNOFILE=500000
StandardOutput=syslog
StandardError=syslog

[Install]
WantedBy=multi-user.target

Save it to /etc/systemd/system/loom.service. Run these to activate it:

sudo systemctl daemon-reload
sudo systemctl start loom.service

You may now inspect the output using:

sudo journalctl -u loom.service

When satisfied everything is running as intended, executing the following will enable the service so that it is started at boot:

sudo systemctl enable loom.service

# Verifying

# Listening Ports

If all is well, you will be able to see these ports opened in each node.

$ sudo netstat -tpnl
Active Internet connections (only servers)
Proto Recv-Q Send-Q Local Address           Foreign Address         State       PID/Program name
tcp6       0      0 :::26656                :::*                    LISTEN      2135/loom
tcp6       0      0 :::46657                :::*                    LISTEN      2135/loom
tcp6       0      0 :::46658                :::*                    LISTEN      2135/loom

# Automation

If combining the configuration files and startup commands seems to be a lot of work, we have a way to automate it using Ansible.

Note that Ansible needs to be installed locally.

The playbook is available here.

You will need to change the inventory to match your nodes and preferred working directory.

Please ensure SSH and sudo access to the nodes are available

# Inventory: inventory.yaml

The inventory specifies the nodes and their IP addresses. If the node only has one IP, use the same for both ansible_host and private_ip. ansible_host is used by Ansible to connect to the host, while private_ip is used by the nodes to communicate with each other.

---
all:
  vars:
    loom_build: build-404
    ansible_ssh_common_args: '-o StrictHostKeyChecking=no'
    working_directory: /home/ubuntu
    user: ubuntu
  hosts:
    loom-0:
      ansible_host: 1.2.3.4
      private_ip: 10.2.3.4
    loom-1:
      ansible_host: 5.6.7.8
      private_ip: 10.6.7.8
    loom-2:
      ansible_host: 4.3.2.1
      private_ip: 10.3.2.1
    loom-3:
      ansible_host: 8.7.6.5
      private_ip: 10.7.6.5

After modifying the inventory with the details of the nodes, execute the playbook:

ansible-playbook -i inventory.yml -vv loom-playbook.yml

# More Automation: Vagrant

There is also a Vagrantfile included to provision a full cluster. Ansible needs to be installed on the host machine.

It is tested with VirtualBox provider. It takes less than two minutes on a decent machine to create and provision 4 nodes.

The following variables may be changed when needed:

# Size of the cluster created by Vagrant
num_instances = 4

# Private Network Prefix
private_network_prefix = "172.31.99."

# Build numbers
loom_build = "build-208"

Note: Vagrant creates its own inventory so inventory.yml is not used.