#Sending messages betwen nodes through global

30 messages · Page 1 of 1 (latest)

tiny tapir
#

I managed to connect nodes and send messages between them, but if i have a lot of nodes that connect and disconnect frequently it seems like the :global module have trouble with network partition

stuck magnet
#

I think you can try to use something like Horde (or Pogo ? or other newer stuff) to handle this kind of stuff!

#

this way you will have a distributed registry 🙂

tiny tapir
#

what about libcluster ?

hybrid oasis
#

Are the other nodes connecting over the internet?

#

If so, using erlang distribution to connect them is grossly insecure

#

...and without tricks, you'll never be able to connect thousands of nodes

tiny tapir
#

yes they are connected over the internet. Which are the alternatives then ?

Sorry for the ignorance, first time trying to build something like this

hybrid oasis
#

Maybe take a step back and explain what you’re trying to build

#

But once a node is connected, it has full access to the node it connected to

#

Quit from what I can tell so far, maybe just use an rpc protocol over tls, like thrift or grpc

tiny tapir
#

What i'm trying to dobuild is a way of monitoring workers. I would like to install an agent on each worker (in my case container) and then from a manager node see all the agents, assign roles, and other stuff

#

Something like an orchestrator

hybrid oasis
#

The worker isn’t primarily a n elixir machine?

#

Who controls the worker?

tiny tapir
#

the worker is a generic container, which can change. The number of workers/container is dynamic and i choose it from swarm

#

docker swarm

hybrid oasis
#

Yes, but who runs them?

#

Do you trust them?

tiny tapir
#

yes

#

i run them

hybrid oasis
#

I'm confused as to why they're connecting over the internet then

tiny tapir
#

probably i missunderstood what you mean "over the internet", sorry english is not my first language. Everything is running on my local network, or aws

hybrid oasis
#

Ah, ok

#

Then why are there so many partitions?

#

(Just asking. Global isn't a great choice for this regardless of partitions)

#

Though global does have customizable ways to handle naming conflicts

#

Also, are the workers special?