Clone Node

Clone node is a configurable node script that when pointed to another instance of HarperDB will create a clone of that instance's config, databases and setup replication. If it is run in a location where there is no existing HarperDB install, it will, along with cloning, install HarperDB. If it is run in a location where there is another HarperDB instance, it will only clone config, databases and replication that do not already exist.

Clone node is triggered when HarperDB is installed or started with certain environment or command line (CLI) variables set (see below).

Leader node - the instance of HarperDB you are cloning. Clone node - the new node which will be a clone of the leader node.

To start clone run harperdb in the CLI with either of the following variables set:

Environment variables

• HDB_LEADER_URL - The URL of the leader node's operation API (usually port 9925).

• HDB_LEADER_USERNAME - The leader node admin username.

• HDB_LEADER_PASSWORD - The leader node admin password.

• HDB_LEADER_CLUSTERING_HOST - (optional) The leader clustering host. This value will be added to the clustering routes on the clone node. If this value is not set, replication will not be set up between the leader and clone.

For example:

HDB_LEADER_URL=https://node-1.my-domain.com:9925 HDB_LEADER_CLUSTERING_HOST=node-1.my-domain.com HDB_LEADER_USERNAME=... HDB_LEADER_PASSWORD=... harperdb

Command line variables

• --HDB_LEADER_URL - The URL of the leader node's operation API (usually port 9925).

• --HDB_LEADER_USERNAME - The leader node admin username.

• --HDB_LEADER_PASSWORD - The leader node admin password.

• --HDB_LEADER_CLUSTERING_HOST - (optional) The leader clustering host. This value will be added to the clustering routes on the clone node. If this value is not set, replication will not be set up between the leader and clone.

For example:

harperdb --HDB_LEADER_URL https://node-1.my-domain.com:9925 --HDB_LEADER_CLUSTERING_HOST node-1.my-domain.com --HDB_LEADER_USERNAME ... --HDB_LEADER_PASSWORD ...

Each time clone is run it will set a value cloned: true in harperdb-config.yaml. This value will prevent clone from running again. If you want to run clone again set this value to false. If HarperDB is started with the clone variables still present and cloned is true, HarperDB will just start as normal.

Clone node does not require any additional configuration apart from the variables referenced above. However, if you wish to set any configuration during clone this can be done by passing the config as environment/CLI variables or cloning overtop of an existing harperdb-config.yaml file.

More can be found in the HarperDB config documentation here.

Note: because node name must be unique, clone will auto-generate one unless one is provided

Excluding database, components and replication

To set any specific (optional) clone config, including the exclusion of any database, components or replication, there is a file called clone-node-config.yaml that can be used.

The file must be located in the ROOTPATH directory of your clone (the hdb directory where you clone will be installed. If the directory does not exist, create one and add the file to it).

The config available in clone-node-config.yaml is:

databaseConfig:
  excludeDatabases:
    - database: null
  excludeTables:
    - database: null
      table: null
componentConfig:
  exclude:
    - name: null
clusteringConfig:
  publishToLeaderNode: true
  subscribeToLeaderNode: true
  excludeDatabases:
     - database: null
  excludeTables:
     - database: null
       table: null

Note: only include the configuration that you are using. If no clone config file is provided nothing will be excluded, unless it already exists on the clone.

databaseConfig - Set any databases or tables that you wish to exclude from cloning.

componentConfig - Set any components that you do not want cloned. Clone node will not clone the component code, it will only clone the component reference that exists in the leader harperdb-config file.

clusteringConfig - Set the replication setup to establish with the other nodes (default is true & true) and set any databases or tables that you wish to exclude from clustering.

Cloning configuration

Clone node will not clone any configuration that is classed as unique to the leader node. This includes clustering.nodeName, rootPath and any other path related values, for example storage.path, logging.root, componentsRoot, any authentication certificate/key paths.

Cloning system database

HarperDB uses a database called system to store operational information. Clone node will only clone the user and role tables from this database. It will also set up replication on this table, which means that any existing and future user and roles that are added will be replicated throughout the cluster.

Cloning the user and role tables means that once clone node is complete, the clone will share the same login credentials with the leader.

Fully connected clone

A fully connected topology is when all nodes are replicating (publish and subscribing) with all other nodes. A fully connected clone maintains this topology with addition of the new node. When a clone is created, replication is added between the leader and the clone and any nodes the leader is replicating with. For example, if the leader is replicating with node-a and node-b, the clone will replicate with the leader, node-a and node-b.

To run clone node with the fully connected option simply pass the environment variable HDB_FULLY_CONNECTED=true or CLI variable --HDB_FULLY_CONNECTED true.

Cloning overtop of an existing HarperDB instance

Clone node will not overwrite any existing config, database or replication. It will write/clone any config database or replication that does not exist on the node it is running on.

An example of how this can be useful is if you want to set HarperDB config before the clone is created. To do this you would create a harperdb-config.yaml file in your local hdb root directory with the config you wish to set. Then when clone is run it will append the missing config to the file and install HarperDB with the desired config.

Another useful example could be retroactively adding another database to an existing instance. Running clone on an existing instance could create a full clone of another database and set up replication between the database on the leader and the clone.

Cloning steps

Clone node will execute the following steps when ran:

1. Look for an existing HarperDB install. It does this by using the default (or user provided) ROOTPATH.

2. If an existing instance is found it will check for a harperdb-config.yaml file and search for the cloned value. If the value exists and is true clone will skip the clone logic and start HarperDB.

3. Clone harperdb-config.yaml values that don't already exist (excluding values unique to the leader node).

4. Fully clone any databases that don't already exist.

5. If classed as a "fresh clone", install HarperDB. An instance is classed as a fresh clone if there is no system database.

6. If clustering is enabled on the leader and the HDB_LEADER_CLUSTERING_HOST variable is provided, set up replication on all cloned database(s).

7. Clone is complete, start HarperDB.

Cloning with Docker

To run clone inside a container add the environment variables to your run command.

For example:

docker run -d \
  -v <host directory>:/home/harperdb/hdb \
  -e HDB_LEADER_PASSWORD=password \
  -e HDB_LEADER_USERNAME=admin \
  -e HDB_LEADER_URL=https://1.123.45.6:9925 \
  -e HDB_LEADER_CLUSTERING_HOST=1.123.45.6 \
  -p 9925:9925 \
  -p 9926:9926 \
  harperdb/harperdb

Clone will only run once, when you first start the container. If the container restarts the environment variables will be ignored.