erigon-pulse/cmd/rpcdaemon

• Introduction
• Getting Started
  • Running locally
  • Running remotely
  • Healthcheck
  • Testing
• FAQ
  • Relations between prune options and rpc methods
  • RPC Implementation Status
  • Securing the communication between RPC daemon and Erigon instance via TLS and authentication
  • Ethstats
  • Allowing only specific methods (Allowlist)
  • Trace transactions progress
  • Clients getting timeout, but server load is low
  • Server load too high
  • Faster Batch requests
• For Developers
  • Code generation

Introduction

Erigon's rpcdaemon runs in its own seperate process.

This brings many benefits including easier development, the ability to run multiple daemons at once, and the ability to run the daemon remotely. It is possible to run the daemon locally as well (read-only) if both processes have access to the data folder.

Getting Started

The rpcdaemon gets built as part of the main erigon build process, but you can build it directly with this command:

make rpcdaemon

Running locally

Run rpcdaemon on same computer with Erigon. It's default option because it using Shared Memory access to Erigon's db - it's much faster than TCP access. Provide both --datadir and --private.api.addr flags:

make erigon
./build/bin/erigon --datadir=<your_data_dir> --private.api.addr=localhost:9090
make rpcdaemon
./build/bin/rpcdaemon --datadir=<your_data_dir> --private.api.addr=localhost:9090 --http.api=eth,erigon,web3,net,debug,trace,txpool

Note that we've also specified which RPC namespaces to enable in the above command by --http.api flag.

Running remotely

To start the daemon remotely - just don't set --datadir flag:

make erigon
./build/bin/erigon --datadir=<your_data_dir> --private.api.addr=0.0.0.0:9090
make rpcdaemon
./build/bin/rpcdaemon --private.api.addr=<erigon_ip>:9090 --http.api=eth,erigon,web3,net,debug,trace,txpool

The daemon should respond with something like:

INFO [date-time] HTTP endpoint opened url=localhost:8545...

Healthcheck

Running the daemon also opens an endpoint /health that provides a basic health check.

If the health check is successful it returns 200 OK.

If the health check fails it returns 500 Internal Server Error.

Configuration of the health check is sent as POST body of the method.

{
   "min_peer_count": <minimal number of the node peers>,
   "known_block": <number_of_block_that_node_should_know>
}

Not adding a check disables that.

min_peer_count -- checks for mimimum of healthy node peers. Requires net namespace to be listed in http.api.

known_block -- sets up the block that node has to know about. Requires eth namespace to be listed in http.api.

Example request http POST http://localhost:8545/health --raw '{"min_peer_count": 3, "known_block": "0x1F"}' Example response

{
    "check_block": "HEALTHY",
    "healthcheck_query": "HEALTHY",
    "min_peer_count": "HEALTHY"
}

Testing

By default, the rpcdaemon serves data from localhost:8545. You may send curl commands to see if things are working.

Try eth_blockNumber for example. In a third terminal window enter this command:

curl -X POST -H "Content-Type: application/json" --data '{"jsonrpc": "2.0", "method": "eth_blockNumber", "params": [], "id":1}' localhost:8545

This should return something along the lines of this (depending on how far your Erigon node has synced):

{
    "jsonrpc": "2.0",
    "id": 1,
    "result":" 0xa5b9ba"
}

Also, there are extensive instructions for using Postman to test the RPC.

FAQ

Relations between prune options and RPC methods

Next options available (by --prune flag):

* h - prune history (ChangeSets, HistoryIndices - used to access historical state)
* r - prune receipts (Receipts, Logs, LogTopicIndex, LogAddressIndex - used by eth_getLogs and similar RPC methods)
* t - prune tx lookup (used to get transaction by hash)
* c - prune call traces (used by trace_* methods)

By default data pruned after 90K blocks, can change it by flags like --prune.history.after=100_000

Some methods, if not found historical data in DB, can fallback to old blocks re-execution - but it require h.

RPC Implementation Status

Label "remote" means: --private.api.addr flag is required.

The following table shows the current implementation status of Erigon's RPC daemon.

Command	Avail	Notes
web3_clientVersion	Yes
web3_sha3	Yes
net_listening	HC	(remote hard coded returns true)
net_peerCount	Limited	internal sentries only
net_version	Yes	remote.
eth_blockNumber	Yes
eth_chainID/eth_chainId	Yes
eth_protocolVersion	Yes
eth_syncing	Yes
eth_gasPrice	Yes
eth_maxPriorityFeePerGas	Yes
eth_feeHistory	Yes
eth_getBlockByHash	Yes
eth_getBlockByNumber	Yes
eth_getBlockTransactionCountByHash	Yes
eth_getBlockTransactionCountByNumber	Yes
eth_getUncleByBlockHashAndIndex	Yes
eth_getUncleByBlockNumberAndIndex	Yes
eth_getUncleCountByBlockHash	Yes
eth_getUncleCountByBlockNumber	Yes
eth_getTransactionByHash	Yes
eth_getRawTransactionByHash	Yes
eth_getTransactionByBlockHashAndIndex	Yes
eth_retRawTransactionByBlockHashAndIndex	Yes
eth_getTransactionByBlockNumberAndIndex	Yes
eth_retRawTransactionByBlockNumberAndIndex	Yes
eth_getTransactionReceipt	Yes
eth_getBlockReceipts	Yes
eth_estimateGas	Yes
eth_getBalance	Yes
eth_getCode	Yes
eth_getTransactionCount	Yes
eth_getStorageAt	Yes
eth_call	Yes
eth_callBundle	Yes
eth_newFilter	-	not yet implemented
eth_newBlockFilter	-	not yet implemented
eth_newPendingTransactionFilter	-	not yet implemented
eth_getFilterChanges	-	not yet implemented
eth_uninstallFilter	-	not yet implemented
eth_getLogs	Yes
eth_accounts	No	deprecated
eth_sendRawTransaction	Yes	remote.
eth_sendTransaction	-	not yet implemented
eth_sign	No	deprecated
eth_signTransaction	-	not yet implemented
eth_signTypedData	-	????
eth_getProof	-	not yet implemented
eth_mining	Yes	returns true if --mine flag provided
eth_coinbase	Yes
eth_hashrate	Yes
eth_submitHashrate	Yes
eth_getWork	Yes
eth_submitWork	Yes
eth_subscribe	Limited	Websock Only - newHeads,
		newPendingTransaction
eth_unsubscribe	Yes	Websock Only
debug_accountRange	Yes	Private Erigon debug module
debug_accountAt	Yes	Private Erigon debug module
debug_getModifiedAccountsByNumber	Yes
debug_getModifiedAccountsByHash	Yes
debug_storageRangeAt	Yes
debug_traceTransaction	Yes	Streaming (can handle huge results)
debug_traceCall	Yes	Streaming (can handle huge results)
trace_call	Yes
trace_callMany	Yes
trace_rawTransaction	-	not yet implemented (come help!)
trace_replayBlockTransactions	yes	stateDiff only (come help!)
trace_replayTransaction	yes	stateDiff only (come help!)
trace_block	Yes
trace_filter	Yes	no pagination, but streaming
trace_get	Yes
trace_transaction	Yes
txpool_content	Yes	remote
txpool_status	Yes	remote
eth_getCompilers	No	deprecated
eth_compileLLL	No	deprecated
eth_compileSolidity	No	deprecated
eth_compileSerpent	No	deprecated
db_putString	No	deprecated
db_getString	No	deprecated
db_putHex	No	deprecated
db_getHex	No	deprecated
erigon_getHeaderByHash	Yes	Erigon only
erigon_getHeaderByNumber	Yes	Erigon only
erigon_getLogsByHash	Yes	Erigon only
erigon_forks	Yes	Erigon only
erigon_issuance	Yes	Erigon only

This table is constantly updated. Please visit again.

Securing the communication between RPC daemon and Erigon instance via TLS and authentication

In some cases, it is useful to run Erigon nodes in a different network (for example, in a Public cloud), but RPC daemon locally. To ensure the integrity of communication and access control to the Erigon node, TLS authentication can be enabled. On the high level, the process consists of these steps (this process needs to be done for any "cluster" of Erigon and RPC daemon nodes that are supposed to work together):

1. Generate key pair for the Certificate Authority (CA). The private key of CA will be used to authorise new Erigon instances as well as new RPC daemon instances, so that they can mutually authenticate.
2. Create CA certificate file that needs to be deployed on any Erigon instance and any RPC daemon. This CA cerf file is used as a "root of trust", whatever is in it, will be trusted by the participants when they authenticate their counterparts.
3. For each Erigon instance and each RPC daemon instance, generate a key pair. If you are lazy, you can generate one pair for all Erigon nodes, and one pair for all RPC daemons, and copy these keys around.
4. Using the CA private key, create cerificate file for each public key generated on the previous step. This effectively "inducts" these keys into the "cluster of trust".
5. On each instance, deploy 3 files - CA certificate, instance key, and certificate signed by CA for this instance key.

Following is the detailed description of how it can be done using openssl suite of tools.

Generate CA key pair using Elliptic Curve (as opposed to RSA). The generated CA key will be in the file CA-key.pem. Access to this file will allow anyone to later include any new instance key pair into the "cluster of trust", so keep it secure.

openssl ecparam -name prime256v1 -genkey -noout -out CA-key.pem

Create CA self-signed certificate (this command will ask questions, answers aren't important for now). The file created by this command is CA-cert.pem

openssl req -x509 -new -nodes -key CA-key.pem -sha256 -days 3650 -out CA-cert.pem

For Erigon node, generate a key pair:

openssl ecparam -name prime256v1 -genkey -noout -out erigon-key.pem

Also, generate one for the RPC daemon:

openssl ecparam -name prime256v1 -genkey -noout -out RPC-key.pem

Now create certificate signing request for Erigon key pair:

openssl req -new -key erigon-key.pem -out erigon.csr

And from this request, produce the certificate (signed by CA), proving that this key is now part of the "cluster of trust"

openssl x509 -req -in erigon.csr -CA CA-cert.pem -CAkey CA-key.pem -CAcreateserial -out erigon.crt -days 3650 -sha256

Then, produce the certificate signing request for RPC daemon key pair:

openssl req -new -key RPC-key.pem -out RPC.csr

And from this request, produce the certificate (signed by CA), proving that this key is now part of the "cluster of trust"

openssl x509 -req -in RPC.csr -CA CA-cert.pem -CAkey CA-key.pem -CAcreateserial -out RPC.crt -days 3650 -sha256

When this is all done, these three files need to be placed on the machine where Erigon is running: CA-cert.pem , erigon-key.pem, erigon.crt. And Erigon needs to be run with these extra options:

--tls --tls.cacert CA-cert.pem --tls.key erigon-key.pem --tls.cert erigon.crt

On the RPC daemon machine, these three files need to be placed: CA-cert.pem, RPC-key.pem, and RPC.crt. And RPC daemon needs to be started with these extra options:

--tls.key RPC-key.pem --tls.cacert CA-cert.pem --tls.cert RPC.crt

WARNING Normally, the "client side" (which in our case is RPC daemon), verifies that the host name of the server matches the "Common Name" attribute of the "server" cerificate. At this stage, this verification is turned off, and it will be turned on again once we have updated the instruction above on how to properly generate cerificates with "Common Name".

When running Erigon instance in the Google Cloud, for example, you need to specify the Internal IP in the --private.api.addr option. And, you will need to open the firewall on the port you are using, to that connection to the Erigon instances can be made.

Ethstats

This version of the RPC daemon is compatible with ethstats-client.

To run ethstats, run the RPC daemon remotely and open some of the APIs.

./build/bin/rpcdaemon --private.api.addr=localhost:9090 --http.api=net,eth,web3

Then update your app.json for ethstats-client like that:

[
  {
    "name": "ethstats",
    "script": "app.js",
    "log_date_format": "YYYY-MM-DD HH:mm Z",
    "merge_logs": false,
    "watch": false,
    "max_restarts": 10,
    "exec_interpreter": "node",
    "exec_mode": "fork_mode",
    "env": {
      "NODE_ENV": "production",
      "RPC_HOST": "localhost",
      "RPC_PORT": "8545",
      "LISTENING_PORT": "30303",
      "INSTANCE_NAME": "Erigon node",
      "CONTACT_DETAILS": <your twitter handle>,
      "WS_SERVER": "wss://ethstats.net/api",
      "WS_SECRET": <put your secret key here>,
      "VERBOSITY": 2
    }
  }
]

Run ethstats-client through pm2 as usual.

You will see these warnings in the RPC daemon output, but they are expected

WARN [11-05|09:03:47.911] Served                                   conn=127.0.0.1:59753 method=eth_newBlockFilter reqid=5 t="21.194µs" err="the method eth_newBlockFilter does not exist/is not available"
WARN [11-05|09:03:47.911] Served                                   conn=127.0.0.1:59754 method=eth_newPendingTransactionFilter reqid=6 t="9.053µs"  err="the method eth_newPendingTransactionFilter does not exist/is not available"

Allowing only specific methods (Allowlist)

In some cases you might want to only allow certain methods in the namespaces and hide others. That is possible with rpc.accessList flag.

1. Create a file, say, rules.json

2. Add the following content

{
  "allow": [
    "net_version",
    "web3_eth_getBlockByHash"
  ]
}

3. Provide this file to the rpcdaemon using --rpc.accessList flag

> rpcdaemon --private.api.addr=localhost:9090 --http.api=eth,debug,net,web3 --rpc.accessList=rules.json

Now only these two methods are available.

Clients getting timeout, but server load is low

In this case: increase default rate-limit - amount of requests server handle simultaneously - requests over this limit will wait. Increase it - if your 'hot data' is small or have much RAM or see "request timeout" while server load is low.

./build/bin/erigon --private.api.addr=localhost:9090 --private.api.ratelimit=1024

Server load too high

Reduce --private.api.ratelimit

Read DB directly without Json-RPC/Graphql

./docs/programmers_guide/db_faq.md

Faster Batch requests

Currently batch requests are spawn multiple goroutines and process all sub-requests in parallel. To limit impact of 1 huge batch to other users - added flag --rpc.batch.concurrency (default: 2). Increase it to process large batches faster.

Known Issue: if at least 1 request is "stremable" (has parameter of type *jsoniter.Stream) - then whole batch will processed sequentially (on 1 goroutine).

For Developers

Code generation

go.mod stores right version of generators, use make grpc to install it and generate code (it also installs protoc into ./build/bin folder).