EventStoreDB 24.6 released

EventStoreDB 24.6 was officially released on the 26th of June 2024. Even though it is a mid-year interim release, there a quite a few feature enhancements plus some exciting new features as well. Keep reading to learn more!

Highlights

New Features

• TCP Client API Plugin
• Apple Silicon for development
• Additional metrics

Feature Enhancements

• Performance enhancements
  
• X.509 authentication in client APIs
  
• Metrics support for Linux, FreeBSD, OSX
• Allow specifying the HTTP status code for health requests
• Fix - Events in explicit transactions can be missing from $all reads
• Miscellaneous enhancements

Breaking Changes

• Persistent subscription checkpoint event type
• Deposed leader truncation: Two restarts needed
• File-copy backup and restore: Two restarts needed
• Unbuffered config setting

TCP Client API Plugin

EventStoreDB v23.10 is the last release to natively support the legacy TCP-based client protocol. Version 24.2 removed it completely.

This release introduces a new plugin that allows to connect to the database with the legacy TCP API. This is a stop-gap solution for customers who have not yet upgraded to the gRPC API.

Read more about enabling and configuration of the TCP API plugin in the documentation.

Apple Silicon for development

An alpha build was available for running EventStoreDB on arm64 processors since EventStoreDB v22.10 using a Docker image.

Since version 24.6, EventStoreDB can now be built from source and executed on macOS with Apple Silicon.

Documentation

Additional metrics

Continuing the work on observability, the following metrics are now available.

• Election counter: eventstore_elections_count
• Projections.
  • Projection status: eventstore_projection_status
  • Percent progress: eventstore_projection_progress
  • Events processed since restart: eventstore_projection_events_processed_after_restart_total
• Persistent Subscriptions:
  • Connection count: eventstore_persistent_sub_connections
  • Total in-flight messages : eventstore_persistent_sub_in_flight_messages
  • Total number of parked messages: eventstore_persistent_sub_parked_messages
  • Oldest Parked Message: eventstore_persistent_sub_oldest_parked_message_seconds
  • Total Number of Item processed: eventstore_persistent_sub_items_processed
  • Last Seen Message:
    • eventstore_persistent_sub_last_known_event_number
    • eventstore_persistent_sub_last_known_event_commit_position
  • Last checkpoint event:
    • eventstore_persistent_sub_checkpointed_event_number
    • eventstore_persistent_sub_checkpointed_event_commit_position

Documentation

Performance enhancements

Several performance and compatibility improvements have been made to EventStoreDB and client SDKs, described below.

Reduced FileHandle usage by 80%

The number of file handles used by the database has been reduced by 80%. On large databases it results in:

• Reduced likelihood of hitting the default OS file handle limit
• Lower overall memory footprint

Faster Startup For Scavenged databases

Large database with thousands of scavenged chunks now starts faster. A 10x improvement has been observed in testing.

Documentation

X.509 authentication in client APIs

Refer to the clients documentation for instructions on how to enable and use X.509 certificates with the EventStoreDB clients.

Metrics support for Linux, FreeBSD, OSX

The following System metrics are now available on the following platforms:

• eventstore_sys_load_avg: Linux, FreeBSD, macOS, Windows
• eventstore_sys_cpu: Linux, FreeBSD, macOS

The following Process metrics are now available on the following platforms:

• eventstore_disk_io_bytes: Linux, Windows and macOS

Allow specifying the HTTP status code for health requests

The HTTP status code to be returned by the /health/live endpoint can be provided using the liveCode query parameter. For example, making a GET HTTP call to /health/live?liveCode=200 will return an empty response with 200 OK status code.

This is useful for liveness probe expecting specific status code.

Fix: Events in explicit transactions can be missing from $all reads

This bug only appears with events written in explicit transactions with the deprecated TCP API. They will not appear on filtered $all reads and subscriptions under specific circumstances.

This fix will also be back ported to v23.10.2 and v22.10.6.

Miscellaneous enhancements

• Index merge will now continue without bloom filters if there is not enough memory to store them.
• Node advertise address is used for identifying nodes in the scavenge log.
• Incomplete scavenges that previously got status Failed will now be marked as Interrupted.
• Error will be logged when the node certificate does not have the necessary usages.
• More information made available in the current database options API response.
• Projections will now log a warning when the projection state size becomes greater than 8 MB.

Persistent subscription checkpoint event type

How to upgrade?

Please refer to the upgrade guide in our documentation.