opsmill/infrahub infrahub-v1.11.10

Infrahub - v1.11.10

on GitHub

Infrahub - v1.11.10 - 2026-10-30

Security

• Updated the base Python version of the Infrahub container image and the development environment from 3.14.7 to 3.14.10, so the pip bundled in the image no longer ships a certifi version affected by GHSA-6q48-6c7w-9gw2.

Added

• Added Redis Sentinel support: when INFRAHUB_CACHE_URL is set to a redis+sentinel:// URL, Infrahub connects to the Sentinel nodes and follows a failover of the Redis master, so a redis-sentinel-proxy is no longer needed. See High availability. (#9449)
• Added the INFRAHUB_BROKER_QUEUE_TYPE setting to declare the durable RabbitMQ queues as quorum queues, so that they stay available when a node of a RabbitMQ cluster stops.

Changed

• Breaking: Raised the Redis requirement to version 6.0 or later, or a Redis-compatible service that supports the HELLO command, because Infrahub now opens every Redis connection with HELLO, which Redis 5.0 and earlier reject. (#9449)
• Changed the error message for a GraphQL query with a syntax error to start with "Query is not valid, ". (#11232)
• Removed the concurrency limit from the clean-up-deadlocks task, so a run that never finishes, for example after a task worker restarts while the run is starting, no longer stops later runs from removing locks left behind by inactive task workers.

Fixed

• Fixed creating or upserting an object whose computed attribute reads a related object given by a human-friendly ID: a related object that does not exist is now reported by name instead of causing an internal server error that clients retried for several minutes, and a related object of a generic kind given with its specific kind is now found. (#11018)
• Fixed repository sync ignoring a changed transformation on an existing artifact definition in .infrahub.yml, which also deleted the artifact definition when its Transformation was renamed. (#11036)
• The Git repository sync, deadlock clean-up and failed-merge detection tasks, which run every minute, no longer store a result in the task manager database and the cache for each run. (#11166)
• Fixed filters, search, and finding an object by name or human-friendly ID on a branch matching objects by a value that the branch had already changed. (#11218)
• Fixed an invalid GraphQL query, a GraphQL query with a syntax error, a webhook rename to an existing name and the deletion of the default IP namespace returning HTTP 500 instead of 422, which made the SDK retry a repository import for 5 minutes before it failed. (#11232)
• Fixed proposed change checks that stayed in progress forever, and blocked the merge, when one of their runs failed or timed out. (#11277)
• Fixed GraphQL queries that request two or more custom root fields in one request, such as two InfrahubResourcePoolUtilization fields, failing with a database read error or hanging.
• Fixed certain schema changes being accepted when data violates them: a schema load or a branch merge is now refused when existing values fall outside a Number attribute's new min_value, max_value or excluded_values, outside a schema-defined number pool's new start_range or end_range, or no longer match a relationship's new peer or a hierarchical node's new parent or children.
• Fixed merges that re-run a Generator set to execute after merge leaving an unused diff stored in the database. Upgrading deletes the diffs that earlier merges left behind.
• Fixed repository sync ignoring a changed class_name on an existing Python Transformation in .infrahub.yml.
• Fixed repository sync ignoring a changed class_name on an existing check definition in .infrahub.yml.
• Fixed repository sync ignoring added, changed or removed targets on an existing check definition in .infrahub.yml, and made the sync fail with an import error when the target group does not exist.
• Fixed task results accumulating in Redis without ever expiring: results now expire two days after they are written.
• Fixed the "Documentation" links in the web interface for built-in object types, such as artifacts, generators, transformations, repositories and accounts: they pointed to documentation pages that no longer exist. Added "Documentation" links to built-in object types that had none, such as check definitions, trigger rules, groups, webhooks, resource pools and permissions. When a schema's documentation value starts with http, the "Documentation" link now opens that URL directly, instead of the documentation home page or a broken link; any other value still opens that path in the documentation that ships with Infrahub. Links to the documentation that ships with Infrahub now end with a trailing slash, so they open the page directly instead of through a redirect that fails when Infrahub runs behind a reverse proxy.
• Fixed the infrahub-testcontainers package starting Neo4j 2025.10.1 instead of 2026.05.0, the version that Infrahub ships with, in test runs of Infrahub Community that do not set NEO4J_DOCKER_IMAGE.
• Fixed the scroll bars of tables, such as object lists, being hidden behind the table's pinned header, footer and actions column, which stopped mouse users from scrolling.

Housekeeping

• The container image build now installs the uv version defined in define-versions.yml (0.13.6) instead of the latest release, and stops when no version is given.