> ## Documentation Index
> Fetch the complete documentation index at: https://docs.canton.network/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> If you are building a Canton Network application, start from https://docs.canton.network/appdev/quickstart.md. To run only a local network for your own Daml code, without building the QuickStart demo, see https://docs.canton.network/sdks-tools/development-tools/localnet.md.

# 3.7.4

> Canton 3.7.4 release notes.

# Release of Canton 3.7.4

Canton 3.7.4 has been released on November 18, 2026.

## Summary

This is a maintenance release that speeds up Ledger API indexer initialization and pruning, and fixes issues with DB connection locks, CantonBFT P2P connections and logical synchronizer upgrades. Its breaking changes affect disabling a DAR, minimum ledger times and the alpha sequencer channels. Participants run a new database migration on startup.

**Summary of breaking changes** (see [Breaking changes](#breaking-changes) for details and migration):

| Ledger API | Admin API and console | Protocol | Configuration |
| :- | :- | :- | :- |
| yes | yes | no | no |

## What's New

### All node operators

* New console command `topology.transactions.export_single_to_file` and new optional `password` parameter of
  `topology.transactions.load_single_from_file`: export and load a signed topology transaction, optionally encrypted
  with a password by the console.

### Participant node operators

* The initialization of the Ledger API indexer after onboarding to a synchronizer needs much less memory and
  completes much faster. In testing, initializing the Ledger API indexer after onboarding a participant to a
  synchronizer with 10 million parties took about 90 seconds after the initial topology import.
* Ledger API pruning is faster. Contract pruning candidates are assembled with more but simpler queries with the help
  of temporary tables to ensure good query plans.
* Previously `canton.remote-participants.<remoteParticipant>.token` was used for both Ledger API and Admin API commands
  in the Canton Console. You can now additionally set `canton.remote-participants.<remoteParticipant>.admin-api-token`
  to use a different token for Admin API commands when Admin API authorization differs from Ledger API authorization.

#### Remote read of a party's ACS (alpha)

* Invoking `get_remote_acs` again with the same arguments while a previous invocation has not
  completed (for example after the previous invocation has failed or its client has disconnected)
  restarts the remote read from the beginning (reusing an existing sequencer channel agreement if
  available). The previous invocation, if still running, fails.
* Cancelling a `get_remote_acs` invocation stops the corresponding ACS replication. The
  remote read can be restarted as described above.
* `canton-builtin-admin-workflow-acs-replication-alpha.dar` is introduced as stable.

#### New database migration on participant startup

A new database index on the Ledger API string interning table that allows looking up interned strings by their value
has been added. The migration runs automatically when the participant starts after the upgrade, so startup takes
slightly longer. The migration time grows linearly with the size of the table. In testing, the migration for
1 million rows took about 2 seconds.

### Synchronizer operators

* The traffic state of members that traffic was purchased for before they joined the synchronizer (protocol version 37
  and up) can now be read back, including the serial needed for the next purchase. Set the new field
  `include_onboarding_members` of `SequencerAdministrationService.TrafficControlState` requests, or the new parameter
  `includeOnboardingMembers` of the sequencer console commands `traffic_control.traffic_state_of_members`,
  `traffic_state_of_members_approximate`, `traffic_state_of_members_at_timestamp`, `traffic_state_of_all_members` and
  `last_traffic_state_update_of_members` to `true`. Both default to `false`, which keeps the existing behavior
  unchanged.

#### Sequencer channel pool size and TTL limits (alpha)

Sequencers with `alpha-sequencer-channel-support` enabled now limit sequencer channels:

* `max-active-sequencer-channels` (default `10`) caps how many channels a sequencer serves at a time.
  When the limit is reached, new channels are rejected with the retryable error
  `SEQUENCER_CHANNEL_POOL_CAPACITY_EXHAUSTED`.
* The new `sequencer-channel-ttl` (default `10m`) closes channels when their TTL expires.
* The new `max-metadata-waiting-time` (default `2m`) closes pending connections that have not sent metadata in time.

Online party replication and `get_remote_acs` retry rejected channels and reconnect automatically when a channel is
closed by the TTL.

## Breaking changes

### Ledger API (application developers)

* **Bounded minimum ledger time**: Commands whose `min_ledger_time_abs` or `min_ledger_time_rel` lies more than
  `canton.participants.<participant>.ledger-api.command-service.min-ledger-time-max-delay` (default 5 minutes)
  ahead of the current ledger time are now rejected with `INVALID_FIELD`. Previously such commands were accepted.
  Applications that request a longer delay must lower it, or you must raise the limit on the participant.

### Participant node operators

* **Alleviate dangerous unvetting on Admin API DAR disable**: To avoid unexpected behavior, the Admin API
  `com.digitalasset.canton.admin.participant.v30.PackageService.UnvetDar` now unvets only the main package of a DAR
  by default. Previously it unvetted all packages in the DAR by default. To get the old behavior, set the new request
  field `UnvetDarRequest.unvet_scope` to `UnvetScope.UNVET_SCOPE_ALL_DAR_PACKAGES`. The console command
  `<participant>.dars.vetting.disable` is adapted accordingly.

### Synchronizer operators

* **Sequencer channels (alpha)**: Sequencers with `alpha-sequencer-channel-support` now serve sequencer channels only
  for synchronizers that run the `dev` protocol version. The limits described in
  [Sequencer channel pool size and TTL limits](#sequencer-channel-pool-size-and-ttl-limits-alpha) are now enforced
  with their default values, so deployments that relied on unlimited or unbounded channels must adjust
  `max-active-sequencer-channels`, `sequencer-channel-ttl` and `max-metadata-waiting-time`.

## Security Fixes

* Dependency updates to address security vulnerabilities: google-cloud-kms 2.105.0 (CVE-2026-119078).

## Compatibility

The following Canton protocol versions are supported:

| Dependency | Version |
| - | - |
| Canton protocol versions | 36, 37 |

Canton has been tested against the following versions of its dependencies:

| Dependency | Version |
| - | - |
| Java Runtime | OpenJDK 64-Bit Server VM (build 21.0.12+2-nixos, mixed mode, sharing) |
| Postgres | Recommended: PostgreSQL 17.11 (Debian 17.11-1.pgdg13+2) – Also tested: PostgreSQL 14.25 (Debian 14.25-1.pgdg13+1), PostgreSQL 15.20 (Debian 15.20-1.pgdg13+1), PostgreSQL 16.16 (Debian 16.16-1.pgdg13+1), PostgreSQL 17.12 (Debian 17.12-1.pgdg13+1), PostgreSQL 18.7 (Debian 18.7-1.pgdg13+2) |
