Create a Ceramic Mainnet Node

[dysbulic]

What would you like to be added?

In order to participate in the Ceramic network, we need to establish a mainnet node (along with an associated IPFS node) at a static IP address, and submit a pull request to Ceramic’s peerlist.

Ceramic relies on DAG-JOSE in IPFS which currently only has a JavaScript implementation.

The IPFS daemon can be run as a subprocess of the Ceramic daemon, but this is less reliable. The ideal setup will be a standalone IPFS daemon with DAG-JOSE which is only accepting connections from the Ceramic daemon. The Ceramic daemon accepts connections from anywhere as users will connect to it directly when updating their profile.

Images are to be stored in web3.storage with only a reference stored in Ceramic, so it isn't necessary for users to connect to the IPFS daemon for those uploads.

Ceramic has published a set of Terraform scripts for deploying to Amazon Web Services. Ideally, it would be nice if we were deploying to Google’s infrastructure since that is where we're currently creating test instances for pull requests, and we already have our toes in three different services: Render, Vercel, & Google.

It may be necessary, however, to use AWS services if we deploy an instance of IdentityLink to verify Discord, GitHub, & Discourse accounts.

Why is this needed?

Creating a mainnet node is part of joining the Ceramic network and required for access to the system.

[polimyl]

i have done some terraform stuff recently, so i could try to translate the aws terraform scripts to google. what's the priority of this?

[dysbulic]

what's the priority of this?

It will need to be done before we can launch Ceramic-backed profiles.

We can point the development instances to the testnet, so it wouldn't be until the changes are ready to be merged into master that we'd have to have it.

I'm guessing two weeks from now assuming no one but me picks up any of the Ceramic migration tickets.

[polimyl]

after having a look at the terraform module and the ceramic docs i think i'll go with just deploying the docker images since the terraform module isn't doing much more. do we need to run ipfs separately or can we use ceramic's in-process ipfs?

[dysbulic]

do we need to run ipfs separately or can we use ceramic's in-process ipfs?

I run a Ceramic daemon locally, and my instance falls over with a "cb is not a function" error every couple days. If we had an out-of-process IPFS node it would be better for data availability.

I'd really like to be running the Go version of the daemon. Supposedly DAG-JOSE is implemented in Go, but they're waiting on IPFS to merge changes to make it possible to deploy it.

[Udit-takkar]

is the data stored on the ceramic node persistent ?

[dysbulic]

is the data stored on the ceramic node persistent ?

Yeah. The node's data is saved to an IPFS node associated with the Ceramic instance. It is responsible for persisting the data.

[polimyl]

ceramicnetwork/peerlist#58

[polimyl]

mainnet peerlist pr has been merged
our ceramic node is now reachable on ceramic.metagame.wtf:7007