To run a graph-node locally in Docker: start docker deamon on your system and run:
docker-compose up
You can then deploy your subgraphs using yarn create-local && yarn deploy-local
When you're done, don't forget to clean up: docker-compose down && rm -rf data
The following procedure will allow you to fork a remote network like Rinkeby or Mainnet, to make some local changes to it and then index it with a local graph-node instance. This is very convenient to debug your subgraph with real network state without the latency and cost of making real transactions.
First, fork a remote network with hardhat:
FORK=true yarn hardhat node --tags none --hostname 0.0.0.0
Run any script you want on this local fork to update its state, e.g:
yarn hardhat run myScript.ts --network localhost
Point the ethereum RPC parameter in docker-compose.yml to localhost, e.g:
ethereum: 'rinkeby:http://host.docker.internal:8545'
Kill, clean and relaunch graph-node containers via docker-compose:
sudo docker-compose down && sudo rm -rf data && sudo docker-compose up
Upload a subgraph to your node, e.g:
cd angle-subgraph-borrow
yarn create-local && yarn deploy-local:rinkeby
That's it, you can now make queries to your subgraph. A visual tool is available at http://localhost:8000/subgraphs/name/angle/sub/graphql
The borrow subgraph has the particularity of updating its entities on every oracle price update. This can be disabled to decrease crash recovery time by setting FAST_SYNC_THRESHOLD to a recent date in constants.ts.
Oracle updates happening before this timestamp will be sampled to compute at most 1 oracle-induced refresh every 3 hours.
Updates happening above this timestamp will be processed normally.
Once crash is recovered, FAST_SYNC_THRESHOLD can be set back to 0, and slower indexing can be relaunched in background to get better historical data.