- Extendable to most Tendermint-based chains
- Support both mainnet and testnet
- Stable playbooks and roles; Customizable variables
- Support essential functions (snapshot, state-sync, public RPC/API endpoints) through separate playbooks
You run one playbook and set up a node.
ansible-playbook main.yml -e "target=juno_main"
ansible-playbook -i inventory_testnet.ini main.yml -e "target=juno_test"
Because we try our best to support the latest node version, it is not recommended for you to sync from Block 1. Rather, please state-sync or start from a snapshot.
For every network where we run a validator on mainnet, we run 3 nodes (Validator, Backup and Relayer). The details of our 3-node infrastructure are documented here.
We have 2 strong opinions about the node configuration:
- Each network will have its custom port prefix. This is to prevent port collision if you run multiple nodes on the same server. For example, Juno's custom port prefix is 126 and that of Osmosis is 125. Since it is rather arbitrary, we are going to force the same convention on you unless you fork the code. The full list of our port prefixes is here. We recommend you follow this conversion.
- Each type of node will have its setting based on Polkachu's "best practice". For example, the main node (Validator) has null indexer, and 100/0/ pruning, and Relayer node has kv indexer and 40000/2000/ pruning. We will force these setting on you unless you fork the code.
Take a look at the inventory.sample
file. You will see an example juno
group with 3 different hosts: juno_main
, juno_backup
, and juno_relayer
. Each host will have the following variables:
ansible_host
: Required. The IP address of the server.type
: Required. It can bemain
,backup
andrelayer
(alsotest
if you are adventurous). Each is opinionated in its configuration settings.prepare
: Optional. If unset, it is default to true. Iffalse
, it will skip setups of firewall, go, cosmovisor, node exporter, promtail, etc. The reason for thefalse
option is because we run many backup/relayer nodes on the same server with setup done already.
Besides the above host variables, you will also specify the following all
variables in the inventory file:
ansible_user
: The sample file assumesubuntu
, but feel free to use other user name. This user need sudo privilege.ansible_port
: The sample file assumes22
. But if you are like me, you will have a different ssh port other than22
to avoid port sniffing.ansible_ssh_private_key_file
: The sample file assumes~/.ssh/id_rsa
, but you might have a different key location.var_file
: It tells the program where to look for the variable file. This is useless for the mainnet, because the var file will automatically be inferred by the network name. However, it is essentially for testnets.user_dir
: The user's home directory. In the sample inventory file this is a computed variable based on the ansible_user. It assumes that it is not a root user and its home directory is/home/{{ansible_user}}
.path
: This is to make sure that the ansible_user can access thego
executable.node_exporter
: Default istrue
. Change it tofalse
if you do not want to install node_exporterpromtail
: Default istrue
. Change it tofalse
if you do not want to install promtaillog_monitor
: Enter your monitor server IP if you install promtail.node_name
: This is your node name for the config.toml file.log_name
: This is the server name for the promtail service.pagerduty_key
: This is the PagerDuty key if you use TenderDuty
One you understand the setup, please first copy it to your own inventory file so you can customize it to suit your needs:
cp inventory.sample inventory
When you are ready install a node, you run:
ansible-playbook main.yml -e "target=HOST_NAME"
Playbook | Description |
---|---|
main.yml |
The main playbook to set up a node |
prepare.yml |
Prepare the server with node exporter, promtail, go, cosmovisor, and firewall rules |
support_backup_node.yml |
Install snapshot, state_sync, resync, genesis and prune script on backup node |
support_snapshot.yml |
Install snapshot script and a cron job |
support_state_sync.yml |
Install state-sync script |
support_resync.yml |
Install weekly scheduled state-sync and recovery script |
support_genesis.yml |
Install a script to upload genesis |
support_prune.yml |
Install a script to prune using cosmprund |
support_public_endpoints.yml |
Set up Nginx reverse proxy for public PRC/ API |
support_seed.yml |
Install seed node with Tenderseed. You need a node_key.json.j2 file so the node_id is consistent |
support_tenderduty.yml |
Install Tenderduty |
support_price_feeder.yml |
Install price feeders for selected networks (such Umee, Kujira, etc) |
ansible-playbook support_seed.yml -e "target=umee_seed seed=190c4496f3b46d339306182fe6a507d5487eacb5@65.108.131.174:36656"
ansible-playbook support_tenderduty.yml -e "target=juno_tenderduty key=junovalcons1qyw2x2sjp40cqasdfyuiahsdfknasdkneafs"
# When you just want to update price feed config
ansible-playbook support_price_feeder.yml -e "target=kujira_main"
# When you just want to update price feed config and service file
ansible-playbook support_price_feeder.yml -e "target=kujira_main price_feeder_password=YOUR_PASSWORD"
# When you just want to update price feed config and service file and binary
ansible-playbook support_price_feeder.yml -e "target=kujira_main price_feeder_password=YOUR_PASSWORD price_feeder_binary=true"
See the group_vars
folder for the list of mainnet and testnet support
All port prefixes come from this repo
In V1, the custom port prefix is 2 digits. However, this hobby project has evolved into a more ambitious one and we have run out of the prefixes. Therefore, V2 introduces a breaking change of the 3-digit custom port prefixes.
If you have a node running based on V1 port prefix system, you do not need to do anything. However, if you are as OCD as Polkachu, you might want to migrate all the previous nodes to comply with the new system. Here is a playbook to manage the migration. You still need to close the old ports that are not longer in use, but this playbook should take care of the rest.
ansible-playbook support_config_update.yml -e "target=juno_main"
Because this repo tries to accommodate as many Tendermint-based chains as possible, it cannot adapt to all edge cases. Here are some known issues and how to resolve them.
Chain | Issue | Solution |
---|---|---|
Axelar | Some extra lines at the end of app.toml | Delete extra lines and adjust some settings these extra lines are supposed to change |
Canto | genesis file needs to be unwrapped from .result.genesis | Unwrap genesis with jq command |
Injective | Some extra lines at the end of app.toml | Delete extra lines and adjust some settings these extra lines are supposed to change |
Kichain | Some extra lines at the end of app.toml | Delete extra lines and adjust some settings these extra lines are supposed to change |
Celestia testnet | inconsistent config.toml file variable naming convention | Manually adjust config.toml file |