docs/managed-datahub/release-notes/v_1_1_0.md
Hotfix release on top of v1.1.3.
25-June-2026
Recommended Versions
Authorization — policy engine performance
Semantic search — enriched dataset bridge documents: Dataset bridge documents now include structured properties, field-level tags and glossary terms, doc links, and domains, improving semantic-search relevance for dataset discovery.
Semantic search — bridge document backfill: The backfill is now significantly faster at production scale (direct full-doc writes, bounded-parallelism batching) and resilient to permanent failures — cursor advances past irrecoverable URNs rather than looping indefinitely.
DATAHUB_USE_OBSERVE_MODELS=false, the smart-assertion trainer no longer imports the observe-models package at module load, eliminating transitive PyTorch and Darts imports and reducing resident memory on slim or low-RAM executor builds.Dataset semantic search is no longer enabled implicitly: Enabling semantic search with the default ELASTICSEARCH_SEMANTIC_SEARCH_ENTITIES=document no longer also bridges datasets; dataset must be listed explicitly (document,dataset). Action: for instances relying on dataset semantic search, set ELASTICSEARCH_SEMANTIC_SEARCH_ENTITIES=document,dataset on GMS and the system-update job — otherwise datasets stop being semantically searchable after upgrade.
Removed REQUEST_MINIMAL_SLACK_PERMISSIONS / requestMinimalSlackPermissions feature flag: This deploy-wide knob hard-excluded the Slack :history scopes from the install manifest. Its job is now done by DATAHUB_SLACK_SERVER_SIDE_HISTORY_ENABLED, which marks the same scopes as optional in the Slack install screen — admins can deselect them per install rather than the choice being baked in at deploy time. If the env var was set on your instance, remove it and either (a) set DATAHUB_SLACK_SERVER_SIDE_HISTORY_ENABLED=true and have the workspace admin deselect the history scopes when refreshing the install, or (b) leave it off, in which case the Slack app will request the full scope set as before.
Subscriptions without explicit notification settings now inherit notification defaults dynamically: GraphQL callers that omit notificationConfig when creating a subscription through syncSubscription, or send notificationConfig without notificationSettings, will leave the subscription configured to use the actor's current notification defaults at delivery time instead of creating a subscription with no sinks. Existing subscriptions are unchanged when notificationConfig is omitted during a sync. Callers that intentionally want a no-sink subscription should send notificationConfig.notificationSettings.sinkTypes: [] explicitly.
MCP tools register_feedback and note_sql_anchor_observation removed: These tools have been replaced by a single note_metadata_observation tool covering both metadata gaps and SQL-anchor quality observations. Agent prompts that reference the old tool names must be updated.
Criterion.values so GMS correctly applies timestamp bounds on smart assertion evaluation queries.Hotfix release on top of v1.1.2. There are no breaking changes or deprecations.
17-June-2026
Recommended Versions
Action Workflows — review modal missing fields — The "View Details" modal for Action Request reviews now shows all fields from the workflow definition (Dataset, Role, Use Case), not just "Requested by" and "Additional Notes".
Ownership — ownerTypes materialization — The ownerTypes aspect is now always materialized from the owners list, preventing stale or missing ownership type data.
Hotfix release on top of v1.1.1. There are no breaking changes or deprecations.
09-June-2026
Recommended Versions
mssql-odbc on the v1.1.0 Wolfi executor, where the driver was missing after the Ubuntu → Wolfi migration.img-src now allows data: and blob: URLs so inline SVG icons render correctly in the frontend after CSP hardening shipped in v1.1.0.This release rolls up hotfixes on top of v1.1.0. There are no breaking changes or deprecations.
08-Jun-2026
Recommended Versions
0) now evaluate correctly._NATIVE_ operator): The Quality/Assertions tab keeps rendering for
_NATIVE_ operators, and _NATIVE_ is now rejected on FieldValues/FieldMetric assertions.WORKFLOW_FORM_REQUEST in the action-request owner source.ssh binary to the executor image (enables SSH-based connections during ingestion/assertion runs).Includes all upstream OSS changes through DataHub Core v1.6.0 — see Updating DataHub for breaking changes.
01-June-2026
severity classification that indicates how significant the failure is. Thresholds can be manually configured through Severity Assignment Rules, or automatically determined by DataHub using signals like deviation from expected bounds, asset importance, and potential downstream impact. SDK tutorial.get_entities now returns aspects on schemaField URNs, and column-level structured properties and documentation are exposed to agents — closing a long-standing gap for column-aware agentic workflows.v1.1.0 inherits breaking changes from upstream OSS v1.6.0. Scan the categories below to see if any apply to your deployment — full details, exact configs, and migration actions are in Updating DataHub — v1.6.0.
Platform & runtime — likely action required for self-hosted, custom plugin, and Compose deployments
DATAHUB_SECRET — secret must be ≥32 bytes or datahub-frontend fails on startup./metrics. Prometheus and Grafana configs need updating.BASE_IMAGE and apkRepositoryUrl build args required.v1.1.0-cloud executor images use Chainguard Wolfi instead of Ubuntu 24.04. Custom Dockerfile layers must use apk (not apt) and a Wolfi-compatible BASE_IMAGE. See Wolfi overview and searching for packages on Wolfi.Auth & API contracts — affects custom integrations
corpUserInfo.active no longer considered for session eligibility.value → values — singular value removed on filter Criterion; custom REST, GraphQL, and SDK clients must send a values array.Ingestion — URN & lineage shifts — may break dashboards, saved searches, and lineage queries keyed on old URNs
DataFlow / DataJob URN shape no longer duplicates the database name.InputFields self-references dropped, column field paths now use warehouse-native names, workbook charts emit warehouse-table edges in ChartInfo.inputs, and DM element schema fixes may surface previously suppressed columns. Redshift users: set default_database / default_schema per connection in the new connection_to_platform_map.Ingestion — config & behavior changes — likely action required for affected recipes
sqlalchemy — recipes with method: ge need pip install 'acryl-datahub[profiling-ge]'. Applies to all SQL connectors including Unity Catalog.incremental_ownership: true / incremental_properties: true to preserve them.extract_policy_tags_from_catalog rewritten on INFORMATION_SCHEMA + batched Data Catalog API; old path removed with no fallback.type = ERROR instead of FAILURE; new severity field (LOW for warn, HIGH for fail). Alerts filtering on type == FAILURE will no longer include infra issues.STRUCTURED_PROPERTIES_DROP_MISSING_PROPERTY_VALUES_WITH_WARNING=false to restore the old behavior.filter_config field renames — recipes using old names will fail.global.datahub.monitoringlangchain-core ≥1.2.28 (CVE-2026-40087)nbconvert 7.17.1 (CVE-2026-39378)python-dotenv ≥1.2.0 (CVE-2026-28684)requests ≥2.33.0 (CVE-2026-25645)langsmith ≥0.7.31 (GHSA-rr7j-v2q5-chgv)mako ≥1.3.11 (CVE-2026-41205)hadoop 3.4.1 (CVE-2024-23454)python-multipart 0.0.26 (CVE-2026-40347)pycurl, GitPython, Jupyter stack, mistune, observe-models (CVE-2026-1839) — minimum versions raisedchat_id / token telemetry linkage; markdown escape sequences handled correctly in conversation titles.runAs restricted to service accounts.get_entities now returns aspects when called on a schemaField URN, and column-level structured properties and documentation are exposed (behind env vars), giving agents richer column context without dataset-level lookups.severity classification that indicates how significant the failure is. Thresholds can be manually configured through Severity Assignment Rules, or automatically determined by DataHub using signals like deviation from expected bounds, asset importance, and potential downstream impact. SDK tutorial.EQUAL_TO operator added.TABLE_STATISTICS metric source added. This runs ANALYZE TABLE ... COMPUTE STATISTICS followed by DESCRIBE TABLE EXTENDED, which reads the cached numRows from the catalog rather than scanning data. On Delta tables this is a metadata-only operation, making it significantly cheaper than a COUNT(*) query on large tables. Read more.PLATFORM_API metric source added. Uses the tables.get API to retrieve the current row count for a Table. This is the most cost-effective option as it does not run any SQL queries or scan any data — the row count is retrieved via a free API call. Read more.These capabilities are in active development and gated behind feature flags. Reach out to your DataHub representative to enable them.
A new workspace for proposing, reviewing, and publishing AI-generated context (documentation, glossary terms, annotations) on assets — with human-in-the-loop governance.
find_sql_context, draft_sql_for_tables, generate_sql_sketch, and note_metadata_observationKeep agent-facing context accurate and consistent. Evals continuously check context quality in DataHub and flag divergent or contradictory documentation before it's published.
INFORMATION_SCHEMAinclude_metric_views in ingestionSql.Databases M-Query supportedqueryinsightscatalog.jsonseverity and improved ERROR vs FAILURE mappingOperator note: Use
WORKER_CONFIG_FETCHER_RETRIESfor alerting on sustained retries. The legacyWORKER_CONFIG_FETCHER_ERRORSgauge is preserved for backwards compatibility but no longer increments on every retry, so any existing alerts on it will silently miss issues.
DATAHUB_SECRET must be ≥32 bytes (see Breaking Changes)./metrics scrape path (see Breaking Changes).// and other encoded URN properties now allowed in URLs.setDomain action failures resolved.