Skip to content

Latest commit

 

History

History
148 lines (104 loc) · 6.06 KB

README.md

File metadata and controls

148 lines (104 loc) · 6.06 KB
Raw

Query node

This is a query-node project generated by hydra-cli. Experiment by modifying schema.graphql and the mappings in the mappings folder, defined in manifest.yml.

For more information about Hydra see: https://dzhelezov.gitbook.io/hydra/

1. Building the node

Run

# Make sure these steps are executed first:
# yarn build; # OR:
# yarn && yarn workspace @joystream/types build && yarn workspace @joystream/metadata-protobuf build
yarn workspace query-node-root build

This will generate event decorations for the mappings via hydra-typegen inside mappings/generated (based on the runtime metadata saved in ../chain-metadata.json) and a graphql-server along with the TypeORM models based on schema.graphql inside generated/graphql-server.

2. Building the mappings

In order to re-build the mappings only, without re-generating the graphql-server code and event decorations (in case there were no changes in the runtime and schema.graphql) you can also run:

yarn workspace query-node-mappings build

3. Initializing the database and starting the node

yarn workspace query-node-root start

This script script will:

  • Bring up joystream-node service (if not already started)
  • Bring up db service (query node is using a PostgreSQL database to store the data)
  • Configure the database (yarn workspace query-node config:dev)
  • Create the database and initialize schema (yarn workspace query-node-root db:prepare && yarn workspace query-node-root db:processor:migrate)
  • Start the GraphQL server service (docker compose up -d graphql-server)
  • Start the Hydra Processor responsible for processing the runtime events and running the mappings' functions (docker compose up -d processor)

4. Stopping the node and removing all associated containers

You can stop the query-node and remove all associated docker containers without affecting any other services by running:

yarn workspace query-node-root kill
  • Database connection settings: DB_NAME, DB_HOST, DB_PORT, DB_USER, DB_PASS
  • Chain RPC endpoint: WS_PROVIDER_ENDPOINT_URI
  • If non-standard types are being used by the Substrate runtime, map type definitions in the json format as an external volume

Follow the links for more information about the indexer service and indexer-api-gateway.

GraphQL Playground assets url

Query node's user interface, GraphQL Playground, is expecting to be served at /graphql. If you are serving the files on path like /query/server/graphql via some nginx proxy, aliasing, etc. you will need to provide the base url to query node server via GRAPHQL_PLAYGROUND_CDN environment variable.

# use the following when serving playground at `/query/server/graphql`
GRAPHQL_PLAYGROUND_CDN="query/server" yarn workspace query-node-root query-node:start:dev

Development

Run integration tests

./tests/network-tests/run-tests.sh

To run tests and keep services alive for further inspection, set PERSIST shell variable to any true-ish value.

PERSIST=true ./tests/network-tests/run-tests.sh

You can then use queries manually in GraphQL Playground (http://localhost:8081/graphql), see docker logs (e.g. docker logs processor), etc.

After running tests in debug mode, you can run more testing scenarios or repeat some. This assumes the scenario is repeatable and any previous test errors didn't break the blockchain or processor state in a critical way.

# run tests first and make sure services stay alive
PERSIST=true ./tests/network-tests/run-tests.sh
yarn workspace network-tests run-test-scenario content-directory

Commenting out some of the scenario's flow calls in network-tests/src/scenarios/content-directory.ts is not relevant to the current scope of development or debugging might speed out the process.

Processor setups

Running processor with local Joystream node and local indexer. It's useful when you want to interact with Joystream node via Pioneer or Atlas and want to check results processed by the processor.

docker compose up -d joystream-node indexer hydra-indexer-gateway processor

# start the GraphQL server and Playground if needed via
docker compose up -d graphql-server

Running processor with remote Joystream node and local indexer. It's useful when you want to synchronize the indexer and processor with Joystream node hosted remotely from scratch. You can analyze any errors in docker logs and tweak mappings.

JOYSTREAM_NODE_WS=wss://target-domain.tmp/ws-rpc docker compose up -d indexer hydra-indexer-gateway processor

Running processor with remote Joystream node and remote indexer. When debugging an error that happened in processor mappings on a remote server that has its own indexer, you can use it and skip potentially time-consuming indexer synchronization

PROCESSOR_INDEXER_GATEWAY=https://target-domain.tmp/query-node/indexer/graphql docker compose up -d processor

Restart processor from the beginning

When debugging an error in mappings that breaks the state and processor needs to be restarted and mappings processed from the beginning, use the following commands.

An example of such error is missed event creating a Member in block 123 and processing block 234 when the said Member created a forum post trying to create. The only way to create a missing Member is to start processing events all over.

docker compose stop graphql-server # ensure graphql server is disconnected from db
docker compose rm -vfs processor # turn off processor
WARTHOG_DB_OVERRIDE=true yarn workspace query-node-root db:reset # reset processor database
docker compose up -d processor # start processor again

Debugging Hydra errors

In situations when an error inside of Hydra occurs but it's not clear what event caused the issue, it might help to add console.log(nextBlock.events) to node_modules/@joystream/hydra-processor/lib/process/MappingsProcessor.js's processBlock function to see what events are being processed.