emqx Files

Enhancements

• #11637 Added extra diagnostic checks to help debug issues when mnesia is stuck waiting for tables. Library Updates: ekka has been upgraded to version 0.15.15, and mria to version 0.6.4.

• #11581 Feature Preview: Planned for EMQX v5.4.0, introducing the concepts of Connector and Action base on data bridge. The existing data bridge will be gradually migrated to Connector and Action. Connector are designed to manage the integration with external systems, while Actions are solely used to configure the data processing methods. Connector can be reused across multiple Actions, providing greater flexibility and scalability. Currently, the migration has been completed for Kafka producer and Azure Event Hub producer.

• The Dashboard now supports MQTT 5.0 publish attribute settings for the rule engine's message republish action, allowing users more flexibility in publishing messages.

Bug Fixes

• #11565 Upgraded jq library from v0.3.10 to v0.3.11. In this version, jq_port programs are initiated on-demand and will not appear in users' processes unless the jq function in EMQX is used. Additionally, idle jq_port programs will auto-terminate after a set period. Note: Most EMQX users are running jq in NIF mode and will not be affected by this update.

• #11676 Hid a few pieces of sensitive information from debug-level logs.

• #11697 Disabled outdated TLS versions and cipher suites in the EMQX backplane network (gen_rpc). Added support for tlsv1.3 on the backplane and introduced new configuration parameters: EMQX_RPC__TLS_VERSIONS and EMQX_RPC__CIPHERS.

The corresponding gen_rpc PR: https://github.com/emqx/gen_rpc/pull/36

• #11734 Fixed clustering in IPv6 network. Added new configurations rpc.listen_address and rpc.ipv6_only to allow EMQX cluster RPC server and client to use IPv6.

• #11747 Updated QUIC stack to msquic 2.2.3.

• #11796 Fixed rpc schema to ensure that client/server uses same transport driver.

• #11798 Fixed the issue where the node could not start after executing ./bin/emqx data import [FILE].

The connection between apikey_key and apikey_name is also enhanced for better consistency and unique identification. - apikey_key: When generating an API key via the dashboard, apikey_key will now create a unique value derived from the provided human-readable apikey_name. - apikey_name Conversely, when using a bootstrap file to generate an API key, apikey_name will be generated as a unique value based on the associated apikey_key.

• #11813 Fixed the schema to ensure that RPC client SSL port aligns with the configured server port. This fix also guarantees that the RPC ports are correctly opened in the Helm chart.

• #11819 Upgraded opentelemetry library to v1.3.1-emqx. This opentelemetry release fixes invalid metrics timestamps in the exported metrics.

• #11861 Fixed excessive warning message printed in remote console shell.

• #11722 Fixed an issue where a Kafka Producer bridge with sync query mode would not buffer messages when in the connecting state.

• #11724 Fixed a metrics-related issue where messages sent to Kafka would be counted as failed even when they were successfully transmitted afterward due to internal buffering.

• #11728 Enhanced the LDAP filter string parser with the following improvements:

• Automatic escaping of special characters within filter strings.

• Fixed a bug that previously prevented the use of dn as a filter value.

• #11733 Resolved an incompatibility issue that caused crashes during session takeover or channel eviction when the session was located on a remote node running EMQX v5.2.x or an earlier version.

• #11750 Eliminated logging and tracing of HTTP request bodies in HTTP authentification and HTTP bridges.

• #11760 Simplified the CQL query used for the Cassandra bridge health check, which was previously generating warnings in the Cassandra server logs.

• #11886 Fixed backward plugin compatibility.

Currently, EMQX validates hook point names, and invalid hook points cannot be used for hook registration. However, some older versions of plugin templates used misspelled hook points, and actual plugins in use may also have this issue. To maintain compatibility with these older plugins, we allow the use of the old hook points for hook registration, but we issue deprecated warnings for them. As before, these hooks will not be called.

• #11897 Fixed the issue of waiting for a loop race condition during node configuration synchronization when cluster nodes are started approximately at the same time.