* Check gateway supported versions
* Fix boxed errors
* Fmt
* feat(client-core): integrate gateway protocol validation into SDK
Enable protocol version checking for all SDK-based clients by using
gateways_for_init_with_protocol_validation in MixnetClientBuilder.
This ensures clients only connect to gateways with compatible protocol
versions, preventing potential communication issues.
- Replace gateways_for_init with gateways_for_init_with_protocol_validation
- Add import for the new validation function
- Protocol validation now active for network monitor and all SDK users
* refactor(client-core): allow gateways with newer protocol versions
Change protocol validation to be more permissive - instead of rejecting
gateways with newer protocol versions, now logs a warning and continues.
This enables graceful degradation when gateways upgrade, relying on their
backward compatibility while signaling users to update their clients.
Changes:
- Accept gateways with protocol version > client version
- Log warning about version mismatch suggesting client update
- Update log messages from "validation" to "check" for clarity
- Add trace logging showing both gateway and client versions
This prevents connectivity issues when gateways upgrade before clients,
improving overall network resilience.
* feat(nym-network-monitor): add diagnostic logging for ConnectionRefused debugging
- Add detailed logging for client rotation lifecycle with [CLIENT_ROTATION_*] tags
- Add request tracking with unique IDs using the random message as identifier
- Add HTTP connection acceptance logging with [HTTP_REQUEST] tags
- Improve locust script with connection error handling and backoff strategy
- Add TCP backlog configuration support (default 1024)
- Add socket configuration with SO_REUSEADDR and configurable backlog
- Add HTTP timeout and tracing layers for better observability
These changes help diagnose when and why ConnectionRefused errors occur during load testing,
particularly around client rotation periods.
* fix(nym-api): calculate route average reliability as node mean instead of route-based
- Changed route average reliability calculation to use mean of all node reliabilities
- Added median calculation alongside mean for better statistical representation
- Fixed null average reliability for old method which doesn't analyze routes
- Rounded all float values to 2 decimal places in API responses for cleaner display
- Store median values in analysis_parameters JSON field
In simulation mode, the nym-api should be able to run performance
simulations regardless of which instance is handling epoch advancement,
since simulations don't perform any blockchain transactions.
This fix:
- Allows simulation mode to proceed when another API is advancing the epoch
- Skips epoch transition attempts in simulation mode
- Skips actual reward distribution blockchain operations in simulation mode
Implements proper socket-level health checking for gateway websocket connections:
- Add is_connection_alive() method to GatewayTransceiver trait
- Implement socket-level health check using TcpStream::peek() on Unix systems
- Add is_gateway_connection_alive() method to MixnetClient for easy access
- Update network monitor to use connection health checks before sending
- Fix axum router state configuration in network monitor
The health check performs actual OS-level socket validation rather than just
checking file descriptor existence, providing reliable connection status.
* Decrease default average packet delay to 15 ms
* Add upgrade for config value
* Fix ip packet router too
* Fix clippy
* Remove message_sending_average_delay from template too
* move SessionType into statsitcis common crate
* add RememberMe to clients config
* change stats collection logic to handle remember me
* set up sdk client to send remember me message
* bump NS API version
* removed support for legacy packet types from NymCodec
I think nodes had plenty of time to upgrade given versioned variant was introduced in 2022
* temp: use local sphinx packet for development
* introduce new messages that use more efficient reply surbs encoding
* checks for incorrect encoding
* generate correct message depending on config value
* fixed current packet version
* made packet type selection configurable
* updated sphinx packet crate to the published version
* fixed wasm build
* fixes in outfox due to sphinx api changes
* additional tests
* clippy
* fixed log/tracing import
* Clean stale partially received buffers
* Tweak timeout
* Do cleanup after handling new messages instead of in the select loop
* Debug logging and remove unreachable
* Downgrade log
* Tweak logs
* tweak whitespace
* Only run the stale check every 10 sec
To reduce the risk of the IPR DoS the client:
- Lower the timeout until the IPR will disconnect a client
- Reduce fewer surbs at a time. Large surb requests increases the
latency until all fragments in the response have been delivered. The
efficiency gains of having large surb requests dimishes quickly for
large sizes as well
Inside client-core we want to prepare the ground for moving a behaviour
close to what we have in the vpn client.
Remove all the recv_with_delay since we want to just stop
Add shutdown condition to all select loops to guard against the shutdown
listener being polled inside the select blocks.
* make packet statistics wasm-compatible
* fixed possible overflow issue in delay controller
* updated wasm-client to be compatible with the current network
* applied same logic to mixfetch client
* removed dead imports
* updated versions
Drazen Urch