Applications & Automation¶

Four installed applications have their stored configuration rewritten by the 2.0.0 migration. If you manage app configuration as code, read this before you replay it.

Search Indexing app configuration fields removed¶

Breaking · Affects: scripted reindex triggers, IaC

- "recreateIndex": {
-   "description": "This schema publisher run modes.",
-   "type": "boolean",
-   "default": true
- },
- "useDistributedIndexing": {
-   "title": "Use Distributed Indexing",
-   "description": "Enable distributed indexing to scale reindexing across multiple servers…",
-   "type": "boolean",
-   "default": true
- },

In 2.0 reindex always recreates the index and distributed indexing is always used. The searchIndexLanguage field's description is also corrected from "Recreate Indexes with updated Language" to "Search index mapping language."

Supporting infrastructure added in the migration:

New diagnostics ship alongside: reindex-drift and index-health checks in /v1/system/validate, plus a cluster-fitness diagnostic (REST endpoint + ops CLI).

RDF Index app defaults changed¶

Behavioural — the migration overwrites your settings

The 2.0.0 migration rewrites RdfIndexApp on installed_apps and apps_marketplace:

The reasoning, from the migration itself: incremental RDF indexing produced unbounded triple growth because relationship-removal paths were not fully reconciled. With a per-run CLEAR ALL the dataset always converges to the relational state; a weekly cadence keeps per-run cost from saturating Fuseki.

The entities reset is a safety measure — with recreateIndex: true issuing CLEAR ALL before indexing, a narrowed entity subset would wipe triples for entity types still present in the database.

RDF configuration defaults¶

Data Insights app: dataQuality module removed¶

— see Data Quality → Data Insights no longer computes data quality.

Because moduleConfiguration is additionalProperties: false, a stored config containing dataQuality prevents DataInsightsApplication from deserialising at startup. The migration strips it from installed_apps, apps_marketplace and every app.version.* row in entity_extension.

MCP application configuration removed¶

· Affects: MCP deployments

UPDATE installed_apps
SET json = JSON_SET(JSON_REMOVE(json, '$.appConfiguration'), '$.allowConfiguration', false)
WHERE name = 'McpApplication';

MCP configuration lives solely in the mcpConfiguration setting. The app-level copy — which no code read — is dropped and the now-empty configure step is hidden in the UI.

Runtime-only fields stripped from stored application data¶

Security-relevant · Affects: anyone reading app JSON

UPDATE installed_apps
SET json = JSON_REMOVE(json, '$.openMetadataServerConnection', '$.privateConfiguration')
WHERE …;
-- and the same for entity_extension rows with extension LIKE 'app.version.%'

openMetadataServerConnection and privateConfiguration were being persisted into application rows and version history. They are runtime-only and are removed. Separately, 2.0 stops exposing the app bot JWT and privateConfiguration secrets in API responses.

Application config union changed¶

· Affects: generated models and strict validators

applicationConfig.json's appConfig oneOf list drops the three Collate AI agent configs and gains a permissive terminal branch:

[
  { "$ref": "external/collateAIAppConfig.json" },
  { "$ref": "external/automatorAppConfig.json" },
  { "$ref": "../../../configuration/slackAppConfiguration.json" },
  { "$ref": "internal/dataInsightsAppConfig.json" },
  { "$ref": "internal/dataInsightsReportAppConfig.json" },
  { "$ref": "internal/searchIndexingAppConfig.json" },
  { "$ref": "internal/cacheWarmupAppConfig.json" },
  { "$ref": "internal/dataRetentionConfiguration.json" },
  { "$ref": "internal/autoPilotAppConfig.json" },
  { "type": "object", "additionalProperties": true }
]

The branch indexes shift — generated unions that key on oneOf position (some code generators do) must be regenerated, not hand-patched.

New tenant-wide app configuration setting¶

settings/settings.json gains settingType: "appConfiguration", backed by api/configuration/appConfiguration.json:

{ "defaultAppMode": "ai | classic | null" }

Seeded from yaml/env on first boot; DB-backed and admin-mutable at runtime afterwards (yaml is ignored once a DB row exists).

Precedence: user preference → persona app mode → tenant defaultAppMode. See UI & Customization → App Mode.

MCP tool contracts¶

Behavioural for MCP clients that parse responses

No tools are removed. 7 tools are added and 15 change.

New tools¶

create_context_memory, find_context, get_asset_context, get_company_context, get_knowledge_content, get_persona_context, search_company_context — all Context Center / AI context tools.

Changed tools¶

create_classification, create_data_product, create_domain, create_glossary, create_glossary_term, create_lineage, create_metric, create_tag, create_test_case, get_entity_lineage, get_test_definitions, patch_entity, root_cause_analysis, search_metadata, semantic_search.

Contract changes that will break clients¶

Other MCP changes¶

• Tools are annotated with title / readOnly / destructive / openWorld hints; create_lineage is marked destructive.
• get_entity_details is authorized against the resolved entity.
• Create tools can set custom properties, and create-overwrites are re-authorized.
• SAML SSO is supported in the MCP OAuth flow; public clients are no longer issued a client secret; OAuth state is echoed without double-encoding; id_token in a URL fragment no longer 400s.
• Content negotiation: application/json is preferred when the client accepts both, and SSE is emitted when the client negotiates text/event-stream.
• Tool-call usage is tracked (mcpToolCallUsage entity, /v1/mcp/usage).
• A server.json is published for the MCP Registry.

CSV import/export becomes a tracked background job¶

· Affects: CSV automation and the UI

background_jobs gains progress, total, result, error, message, cancelRequested and completedAt; a new background_job_logs table stores per-job log lines. jobType gains CSV_IMPORT, CSV_EXPORT and AUDIT_EXPORT; status gains CANCELLED.

New API:

In 1.13 async CSV progress was delivered only over WebSocket — a client without a live socket had no way to observe the job. 2.0 makes jobs pollable and downloadable, and the UI surfaces them in a Background jobs tray.

Async entity restore is also available: PUT /v1/{entityType}/restore?async=true returns 202 with a job id and notifies over the RESTORE_ENTITY_CHANNEL WebSocket channel.

Entity metrics gain full CSV support: GET /v1/metrics/name/{name}/export, /exportAsync, PUT /v1/metrics/name/{name}/import, /importAsync, GET /v1/metrics/documentation/csv.

Concurrency limits for background fan-out¶

· New openmetadata.yaml block

asyncOperations:
  maxConcurrentDbTasks: ${ASYNC_MAX_CONCURRENT_DB_TASKS:-25}
  maxConcurrentRdfWrites: ${ASYNC_MAX_CONCURRENT_RDF_WRITES:-8}
  dataInsightsMaxConcurrentDbTasks: ${DATA_INSIGHTS_MAX_CONCURRENT_DB_TASKS:-16}

Prevents background fan-out from exhausting the request connection pool. Tune down if you see connection-pool starvation during reindex or Data Insights runs.

Search & vector infrastructure¶

Cache¶

New admin endpoints: GET /v1/system/cache/keys, POST /v1/system/cache/invalidate, POST /v1/system/cache/invalidate/entity.

The cache block documentation changes from "Default: Disabled (uses NoopCacheProvider)" to "Default: none (no external dependency). Set CACHE_PROVIDER=redis to enable the Redis L2 cache; it falls back to NoopCacheProvider if Redis is unreachable at startup." The default is unchanged.