feat: Add deployment option for etherlink

[puppetninja]

example for values.yaml

node_config_network:
  chain_name: ghostnet
images:
  octez: 'tezos/tezos:octez-v20.2'
nodes:
  rolling-node:
    storage_size: 100Gi
    runs:
      - octez_node
    instances:
      - is_bootstrap_node: false
        config:
          shell:
            history_mode: rolling
          metrics_addr: ["0.0.0.0:9932"]
    node_selector:
      kubernetes.io/hostname: 'scw-midlpr-sclw-par-bm-pool-pensive-jan-24a9b6'
    storageClassName: 'local-storage'
rolling_snapshot_url: 'https://snapshots.eu.tzinit.org/ghostnet/rolling'
snapshot_source: null
prefer_tarballs: false
smartRollupNodes:
  etherlink-ghostnet-observer-0:
    mode: observer
    rollup_address: sr18wx6ezkeRjt1SZSeZ2UQzQN3Uc3YLMLqg
    snapshot_url: https://snapshots.eu.tzinit.org/etherlink-ghostnet/eth-ghostnet.archive
    preimages_endpoins: https://snapshots.eu.tzinit.org/etherlink-mainnet/wasm_2_0_0
    node_rpc_url: http://rolling-node-0.rolling-node:8732
    nodeSelector:
      kubernetes.io/hostname: 'scw-midlpr-sclw-par-bm-pool-pensive-jan-24a9b6'
    storageClassName: 'local-storage'

[drchrispinnock]

You might want to use a full node and not a rolling node, otherwise initially recovery will not be good.
Also you have eu hardcoded for the snapshot. Would be good to do some region detection.

[puppetninja]

You might want to use a full node and not a rolling node, otherwise initially recovery will not be good. Also you have eu hardcoded for the snapshot. Would be good to do some region detection.

This is a POC values.yaml to deploy a octez-node + octez-rollup-node + octez-evm-node stack on k8s. Yes in prod we can deploy a full octez-node for better recovery.

As for snapshot url this is a value passed to chart, I think it would be better to let user determine the url instead of making decision for them. I can add some comments in the default values.yaml to point to tzinit url for better documentation of etherlink snapshot.

[drchrispinnock]

Why are you using --devmode?

[drchrispinnock]

How are you seeding the EVM directory? From a static snapshot?

[puppetninja]

Why are you using --devmode?

this is what is in prod for ghostnet now, and we plan to be in consistent with the prod

[puppetninja]

How are you seeding the EVM directory? From a static snapshot?

It is build on top of the rollup state

https://github.com/tacoinfra/tezos-k8s/pull/16/files#diff-21923754c42e444f7dddac36a596eaa0fc750e5f104be50b10cccc50ad7a296aR10

[drchrispinnock]

From the core dev team on --devmode

"The argument is kept for backward compatibility but is being ignored by the node"

What is the perceived difference?

On the rollup state - is the rollup node down at the time you do the evm-node initialisation? If not, it probably needs to be.

[puppetninja]

From the core dev team on --devmode

"The argument is kept for backward compatibility but is being ignored by the node"

What is the perceived difference?

Okay, I will delete it

On the rollup state - is the rollup node down at the time you do the evm-node initialisation? If not, it probably needs to be.

the rollup is not in shutdown state, but when evm init from rollup, it would try to acquire the lock first, rollup will be in waiting state when evm acquires the lock, I double checked with dev team and got the confirmation.

[drchrispinnock]

I'm surprised they confirmed on this. I will double check because I got the opposite answer.