You are viewing a single comment's thread from:
RE: SteemPower.org & 'charlieshrem' Witness Update
I'm not sure if there are still something wrong in your infrastructure. Anyway here are my humble suggestions:
- Primary block producing node: run "official" "low_mem_node" compiled in "release" mode on a dedicated (no other duty) VPS or server with 8GB RAM (no swap needed) and good network connection. Remove "account_history" plugin from "config.ini".
- Secondary/backup node: it's OK to use seed node or node with other duties as a backup. But if need to restart it frequently, better have another node running as 2nd backup.
- Use different signing keys on each node.
- The fail-over script can connect to a cli_wallet which connects to public API servers e.g. steem.ws, don't have to connect to your own nodes. It's OK to use your backup nodes if you can make sure they're running when needed.
Hope that you'll have better performance in the future.
Hi there, thanks for the tips.
To be clear, each node does of course use different signing keys and the websocket proxy I wrote actually attempts public API servers if our own servers are down.
With regards to the witness node being low memory: this might in fact make sense but i'm loathe to touch the configuration right now until more hardware becomes available - which should be within a week.
At that point i'll look at swapping the primary witness node to low memory on a dedicated VPS. The hetzner server can then be used more freely for R&D without causing missing blocks.