xrpscan / hermes Goto Github PK

Looking for a command to reclaim disk from Hermes.

In case of rippled connection failure due to network, rippled upgrade or other outages - Hermes does not automatically reconnect to its rippled. It should auto-reconnect because not missing any validation messages is the entire premise of Hermes.

Config option SERVER_HOSTNAME should be used to issue certificates for now. Support for multiple domains via subjectAltName may be added later.

Requested by Rabbit: Can you add a setting to disable TLS certificate verification, so I can connect to a rippled node that uses a self-signed cert?

npmlog supports the following log levels:

log.silly(prefix, message, ...)
log.verbose(prefix, message, ...)
log.info(prefix, message, ...)
log.http(prefix, message, ...)
log.warn(prefix, message, ...)
log.error(prefix, message, ...)

Add support for setting custom log level in .env file.

.env.example config file recommends RIPPLED_URL = 'ws://localhost:6006' as the default RIPPLED_URL. Port 6006 is the admin port, and fetching validation messages via the admin port is an overkill. Change the recommended url to use the unprivileged 6005 port.

Hermes can show a helpful page at its root url /. When Hermes http service is proxy passed behind nginx/apache, the root url can show info about auto-configure url, peer list etc.

Validation.create emits MongoServerError due to unique index on validation_public_key+ledger_index. Change log level for these errors to verbose since they're mostly inconsequential.

Seems the newer LTE Ubuntu 22.04.1 LTS uses libssl3, and thus libssl1.1 is deprecated at this point.

npm run peer [add|ls|remove] or yarn peer [add|ls|remove] don't return control back to the shell. This is probably happening because database connection is still active. Fix database connection logic so the connection can be closed when needed.

yarn keypair command should list its subcommands instead of returning quietly. Something like:

yarn run v1.22.19
$ node dist/cmd/keypair/index.js help
index.js [command]

Commands:
  index.js generate     Generate a new keypair and save      [aliases: gen, new]
  index.js public       Print public key                  [aliases: pub, pubkey]
  index.js cert         Print X.509 certificate     [aliases: certificate, x509]
  index.js fingerprint  Print certificate fingerprint              [aliases: fp]

Options:
  --version  Show version number                                       [boolean]
  --help     Show help                                                 [boolean]
Done in 0.10s.

Update gRPC message types and add requesting node/host information in it. Additionally, this information should also be logged so the operators can review it.

As hermes processes validations, it should build a local list of validators and serve it via an API endpoint.

Add a standalone yarn tsc script that invokes tsc.

The code that auto-appends /api/v1/server/config should handle URLs with trailing slashes correctly. In this example:

npm run peer add https://hermes.rippleitin.nz:50588

works, but:

npm run peer add https://hermes.rippleitin.nz:50588/

throws 404 error.

Handle these errors thrown by gRPC client connections:

[2] node:events:515
[2]       throw er; // Unhandled 'error' event
[2]       ^
[2] 
[2] Error: 14 UNAVAILABLE: read ECONNRESET

Hermes stores a large number of validation messages locally in it database. There's no easy way to verify signatures for validation messages stored in the db. A command that iterates over the validations and verifies them all would be handy.

npm run validation verify --all
npm run validation verify --from LEDGER_FROM --to LEDGER_TO

FATAL ERROR: Ineffective mark-compacts near heap limit Allocation failed - JavaScript heap out of memory

JS heap out of memory error is thrown when the delta between ledgerIndexMin and ledgerIndexMax is large. Review Mongoose driver code.

Requested by Rabbit - One standalone hermes instance connects to a single rippled server only. Sourcing validations from multiple rippled servers can provide better visibility of the network.

setInterval(pollAllPeers, ENV.PEERSYNC_POLL_INTERVAL, peerManager) in peersync processor would keep starting new connections, even if the previous one is active. Added a lock to ensure there's only one client connection per peer.

This error is replicated in the Hermes environment using CentOS 8. This has also been a reported error with previous Node versions

Fix that was applied and works (for my environment).

Stop PM2

Edit file: /etc/systemd/system/pm2-root.service

Add new line: Environment=PM2_PID_FILE_PATH=/run/pm2.pid

And replace: PIDFile=/root/.pm2/pm2.pid to: PIDFile=/run/pm2.pid

restart PM2

Source: https://stackoverflow.com/questions/62814539/pm2-keeps-getting-killed-every-90-seconds-on-centos-8