Skip to content

Latest commit

 

History

History
120 lines (84 loc) · 4.91 KB

README.md

File metadata and controls

120 lines (84 loc) · 4.91 KB
Raw

read-rpc-server

Create .env file in the project root or modify the existing configuration file.

See configuration here.

AWS_ACCESS_KEY_ID=ACCESS_KEY
AWS_SECRET_ACCESS_KEY=SECRET_ACCESS_KEY
AWS_DEFAULT_REGION=eu-central-1
AWS_BUCKET_NAME=buket_name
NEAR_RPC_URL=https://rpc.testnet.near.org
DATABASE_URL=127.0.0.1:9042
DATABASE_USER=username
DATABASE_PASSWORD=password
SERVER_PORT=8888
RUST_LOG=info

Available RPC methods

See the NEAR RPC API for the full list of available methods.

Table of rpc method states here.

Custom rpc methods here.

Compile

$ cargo build --release

Run

Logging

To run an instance of Jaeger, you can use Docker:

$ docker run -d -p6831:6831/udp -p6832:6832/udp -p16686:16686 -p14268:14268 jaegertracing/all-in-one:latest

Run project with command to get metrics

cargo run --release --features tracing-instrumentation

You can visit the Jaeger page by going to http://localhost:16686

Run RPC service

$ ./target/release/read-rpc-server

NEAR RPC API

https://docs.near.org/api/rpc/introduction

Feature flags

send_tx_methods (default: true)

This feature flag enabled the methods for sending transactions to the network. This includes the following methods:

  • broadcast_tx_commit
  • broadcast_tx_async

Might be useful to disable if you want to run a totally read-only node that doesn't even proxies transactions to the network.

Note the methods will be still available in the API, but will return an error if the feature flag is disabled.

tracing-instrumentation (default: false)

This feature flag enables the tracing instrumentation for the RPC server. See the Logging section for more details.

scylla_db (default: true)

See the documentation here

scylla_db_tracing (default: false)

This feature flag enables the tracing instrumentation for the ScyllaDB client (database crate). Database tracing means additional debug-level logs for all the database queries.

shadow_data_consistency (default: false)

This feature flag enables the shadow data consistency checks. With this feature on, all the queries to the read-rpc-server will be executed twice: once against the main database, and once against the real NEAR RPC. If the results are different, warning-level logs are emitted. Note we've decided that read-rpc-server will return its own response if the results are incorrect to make the debugging easier.

Warning This feature is created for the early-stage of lunching the read-rpc-server to catch all the bugs and inconsistencies. You don't need this feature if you are not a contributor to the read-rpc-server.

account_access_keys (default: false)

We encountered a problem with the design of the table account_access_keys and overall design of the logic around it. We had to disable the table and proxy the calls of the query.access_key_list method to the real NEAR RPC. However, we still aren't ready to get rid of the code and that's why we hid it under the feature flag. We are planning to remove the code in the future and remove the feature flag.

Redis (Optional)

Redis is used as an optional dependency when paired with the near_state_indexer feature. To use Redis, you need to specify the REDIS_URL in the .env file and set up a server using docker or redis-server command. Redis can be used to cache the state data indexed by the near_state_indexer for faster retrieval and reduced load on the database.

Metrics (Prometheus)

The read-rpc-server exposes Prometheus-compatible metrics at the /metrics endpoint.

All the metrics are defined in src/metrics.rs file. Two main categories of metrics are:

  • Total number of requests for a specific method
  • Number of errors for a specific method (works only if the shadow_data_consistency feature flag is enabled)

The latter is split by the suffix code to differentiate the reason for the error when possible:

  • 0 - ReadRPC returns a Success result, and NEAR RPC returns a Success result, but the results don't match
  • 1 - ReadRPC returns Success result, and NEAR RPC returns an Error result
  • 2 - ReadRPC returns an Error result, and NEAR RPC returns Success result
  • 3 - ReadRPC returns an Error result, and NEAR RPC returns an Error result, but the results don't match
  • 4 - Could not perform consistency check because of the error (either network or parsing the results)

For example, method block will have these metrics:

  • BLOCK_REQUESTS_TOTAL
  • BLOCK_ERROR_0
  • BLOCK_ERROR_1
  • BLOCK_ERROR_2
  • BLOCK_ERROR_3
  • BLOCK_ERROR_4