You can create named Agents that can be called from any connected node in a cluster. This technique can be helpful when you have a cluster, don't necessarily know which node will handle the work and you want to ensure consistency between nodes by avoiding duplication between nodes. If you don't, you can run into a situation where a you have multiple Agents with different states in different nodes.
If you ever choose to use this technique, consider using a Supervision tree that tries to keep its state across nodes. If one of your Agents goes down, you'd probably want to handle that somehow. There are libraries that currently support this functionality. On to the example:
Create two named `iex` instances in two separate terminal instances:
# On terminal 1:
iex --name t1@127.0.0.1
# On terminal 2:
iex --name t2@127.0.0.1
# Connect instances together. From either t1 or t2, connect to the other
# instance:
# from t2
Node.connect(:"t1@127.0.0.1") #=> true
# Assert we are connected. Check from either instance:
# from t2
Node.list() #=> [:"t1@127.0.0.1"]
# from t1
Node.list() #=> [:"t2@127.0.0.1"]
# Now, let's create an Agent from from `t1`:
Agent.start_link(fn -> %{t1: true, t2: false} end, name: {:global, "choose a name"})
# from `t2`
Agent.get({:global, "choose a name"}, & &1) #=> %{t1: true, t2: false}
## Visual
|--------| |--------|
| t1 |----| t2 |
|--------| |--------|
| | | |
| At1 | | |
| | | |
|--------| |--------|
- At1 -> `global` named Agent
`t2` can communicate with `At1` in a connected BEAM cluster
Thanks for reading, PDG
Home