Continuous Deployment GitOps with Kubernetes

• Spoke with Marcus (Ognots) from NetOps team
• Will be an alpha user of their Gitops process in 1 - 2 months
• They are currently writing run books to make things easy for us
• GitOps using Weave
• Our medusa deployment will be authorized based on Github team permissions
• Need to create templates / helm chart for deploying Medusa
• Writing a Kubernetes Operator will be helpful for partners to easily deploy Medusa with Kubernetes
• We can run Medusa in a namespace under a shared PL Kubernetes cluster
• We can also run Medusa in our own cluster if we need to

Jonathan participating in Launchpad

• Likely 1 - 3 hours per day for 6 weeks
• Learn in-depth technical details about IPFS, IPLD, libp2p, Filecoin from team leads
• Meet others in PLN

Decision

Not worth it considering:

• Uncertainty after 6 month contract
• Amount of work to be done for Medusa

Actually we did a U-turn after a long discussion, with the key caveats:

(1) accept that you may not finish everything in a day or for the whole launchpad, that you may not have time to read everything you wanted, that a code snipped is still not working but that you need to stop and go back to Medusa. i.e seeing this as a

very low engagement /   low expectation commitment

(2) be honest with yourself and Nicolas about saying when things get out of hand, and you’ve spend more than alloted time per day.

Reencryption Refactor PR

https://github.com/medusa-network/medusa/pull/54

• Genericity of Suite has to be tied to concrete triple (CURVE, ENCRYPTION, HASH)
• Enums to match packet types to state machine (Concrete message —> State machine)
• Downcast / dynamic dispatch to go from generic to concrete (State Machine —> Concrete message)
• Oracle Smart Contract designated by id —> State machine (i.e. id = "BN254_KEYG1_HGAMAL”)

Hanging Tests

Embed GitHub

Decision

• Continue debugging rendezvous test until Monday and then update Github Issue with findings
• Factory / Oracle tests do not use a rendezvous server to connect to peers
• Update ethers-rs, libp2p, tokio dependencies

Rendezvous / setup_network / libp2p

One or more nodes wait to establish connection with all other nodes

ideas:

• Registration TTL expires for a node before all other nodes connect to it

Orthogonal fixes:

1. Do not attempt to connect to self
2. Do not register rendezvous server address as a valid address for other nodes

Logs for critical section of peer discovery

• See node with 1 peer at top of logs
• See rdv server send 4 peers to all 4 nodes
• See node still with 1 peer at bottom of logs

Hanging Anvil / Provider