Skip to main content

Open Data Contract Standard

Overview​

The Open Data Contract Standard (ODCS) is the Linux Foundation / Bitol open YAML standard for data contracts. It defines a vendor-neutral schema covering a contract's fundamentals, schema, data quality rules, servers, ownership, tags, SLA, support, and custom properties. Learn more at the audience-friendly Bitol home page and the canonical open-data-contract-standard repository (spec text and JSON Schemas).

ODCS v3 describes a producer-published dataset specification — what a dataset should look like — rather than a bilateral agreement between a specific producer and consumer. The DataHub odcs source therefore models each contract as a logical dataset on the odcs platform (a Logical Model), not as a dataContract entity. A single ODCS file may describe multiple tables: each schema[] entry becomes its own logical odcs Dataset carrying dataset properties, canonical schema metadata, ownership, tags, a link to the source document, one Assertion per quality[] rule, and one schema-compliance assertion — all attached to the logical dataset. When a schema[] entry resolves to a physical dataset (derived from the contract's typed servers[]), the source also links the physical dataset to the logical one with logicalParent (the PhysicalInstanceOf relationship). Propagation of the contract's expectations onto physical instances is handled by DataHub through that relationship — the source itself never writes assertions against physical datasets.

Logical Models are in private beta

Logical Models render in the DataHub UI only when the LOGICAL_MODELS_ENABLED feature flag is on (off by default). The logical odcs datasets — and the assertions attached to them — are ingested either way, but are not displayed until the flag is enabled. See Limitations on the module page for the recommended workflow.

Concept Mapping​

The table below lists the ODCS fields covered. Anything not in this table is deliberately out of scope — see Limitations on the module page.

Source ConceptDataHub ConceptNotes
id + schema[].nameLogical odcs Dataset URNOne logical dataset per schema[] entry. URN name defaults to {contract_id}.{schema_name} (configurable via logical_dataset_name_template).
name + schema[].namedatasetProperties.nameDisplay name "<contract.name> — <schema.name>" when both are present.
id, version, apiVersion, status, domain, …datasetProperties.customProperties (odcs.*)Provenance keys: odcs.id, odcs.version, odcs.apiVersion, odcs.status, odcs.schemaName, odcs.physicalName, odcs.sourceFile, odcs.qualityRuleCount.
schema[].description (fallback: contract description)datasetProperties.descriptionPer-table description wins; description.purpose / usage / limitations objects are concatenated.
schema[].properties[] (name, logicalType, physicalType)schemaMetadata.fields[]Canonical schema on the logical dataset. Types map to a SchemaFieldDataType; unmapped types fall back to NullType and are reported.
Property required / primaryKeyschemaField.nullable / isPartOfKeynullable = not required. Nested properties[] use dotted field paths (address.city).
Property description / tags[]schemaField.description / globalTagstag_prefix is prepended to tags if configured.
Top-level tags[]globalTags on the logical datasetEmitted when replicate_contract_metadata=true (default).
team[] entries (role, username)ownership on the logical datasetrole maps to an OwnershipType; usernames/emails map to corpUser owners. Emitted when replicate_contract_metadata=true.
authoritativeDefinitions[] (root, schema, and property level)institutionalMemoryAuthor-provided URLs linked from the logical dataset.
Typed servers[] entrylogicalParent on the physical datasetThe PhysicalInstanceOf relationship from physical to logical. Platform derives from the spec-required servers[].type; names are fully qualified per platform.
schema[].properties[] (entire declared schema)SchemaAssertionInfo (DATA_SCHEMA assertion) on the logical datasetPins the contract's schema so drift is an evaluable violation. Compatibility defaults to SUPERSET (configurable).
v3.1 metric: nullValues with mustBe: 0FieldAssertionInfo + FieldValuesAssertion (NOT_NULL)Targets the logical dataset.
v3.1 metric: nullValues with another thresholdFieldAssertionInfo + FieldMetricAssertion (NULL_COUNT / NULL_PERCENTAGE)unit: percent selects the percentage metric.
v3.1 metric: duplicateValues / v3.0 rule: duplicateCount with mustBe: 0FieldAssertionInfo + FieldMetricAssertion (UNIQUE_PERCENTAGE == 100)Zero duplicates ⇔ fully unique column.
v3.1 metric: invalidValues / v3.0 rule: validValuesFieldAssertionInfo + FieldValuesAssertion (IN or REGEX_MATCH)Allowed values from arguments.validValues (or the v3.0 direct validValues list); arguments.pattern maps to REGEX_MATCH.
metric: rowCount (v3.0 + v3.1)VolumeAssertionInfounit: rows only; percent-based row counts are preserved as custom assertions.
type: sql with a query and a mappable thresholdSqlAssertionInfoThe query's metric is compared with the mustBe* operator.
All other quality[] rulesCustomAssertionInfoOriginal rule intent preserved verbatim as logic. See Quality rule mapping.

Module odcs​

Incubating

Important Capabilities​

CapabilityStatusNotes
Descriptions✅Dataset and column descriptions.
Detect Deleted Entities✅Via standard stateful ingestion (stateful_ingestion.remove_stale_metadata): only the logical odcs Datasets and Assertions ODCS owns are stale-removed; physical datasets and their logicalParent links are never marked removed.
Extract Ownership✅Owners derived from team[].
Extract Tags✅Top-level and column-level tags.
Schema Metadata✅Canonical schema (types, descriptions, keys) from schema[].properties[].

Overview​

The odcs module ingests Open Data Contract Standard (ODCS) v3.0 and v3.1 YAML files from a path, directory, or glob, and models each contract as a logical dataset on the odcs platform: dataset properties, canonical schema metadata, ownership, top-level and column-level tags, a link to the source document, and the contract's quality[] rules as Assertions attached to the logical dataset. When a schema[] entry resolves to a physical dataset (derived from the contract's typed servers[]), the source also emits a logicalParent link from the physical dataset to the logical one. ODCS is governed by the Linux Foundation under the Bitol project; see bitol.io and the open-data-contract-standard repository.

Looking for the deprecated *.dhub.dc.yaml CLI?

The datahub datacontract upsert flow that consumes *.dhub.dc.yaml files is unrelated to this source and is being phased out. To ingest ODCS YAML, use source.type: odcs via datahub ingest -c <recipe.yml> instead.

Prerequisites​

A single ODCS file may describe multiple tables. Each schema[] entry becomes its own logical odcs Dataset. Contract-level metadata (description, owners, top-level tags) applies to every logical dataset by default. See the Concept Mapping table on the platform Overview above for per-row semantics, and the Limitations section below for how to disable replication.

  • ODCS v3.0 and v3.1 are supported (any 3.0.x patch level validates against the same v3.0.2 JSON Schema; 3.1.0 against the v3.1 schema). Contracts whose apiVersion reports v2.x — or any value outside odcs_versions — are skipped with a warning.
  • Assertions always attach to the logical dataset. Quality rules and the schema-compliance assertion are emitted whether or not a physical table exists yet, so contract-first workflows keep their expectations. Propagation of those expectations onto bound physical datasets is handled by DataHub via the PhysicalInstanceOf relationship — not by this source.
  • Physical binding is derived from the contract itself. The spec requires servers[].type; the source maps supported types (postgres, mysql, snowflake, bigquery, redshift, databricks, sqlserver, trino) to DataHub platforms and composes fully-qualified dataset names from the server's own fields (e.g. database.schema.table). Use server_overrides to refine env / platform_instance, or physical_urn_overrides (keyed by contract id, then schema[] entry name) for explicit URNs. Binding affects only the logicalParent link.
  • Derived physical URNs are verified by default. With a DataHub graph available (datahub-rest sink), a derived URN that does not exist in DataHub is left unbound with a warning instead of creating a stub dataset (verify_physical_urns_exist: false opts out). With a file sink there is no graph, and links are emitted without verification.
  • Logical Models are in private beta. The logical odcs datasets this source emits — and the assertions attached to them — render in the UI only when LOGICAL_MODELS_ENABLED is enabled (off by default). The metadata is ingested either way. The recommended workflow is: ingest the physical platform source (postgres / snowflake / …), ingest ODCS, and enable LOGICAL_MODELS_ENABLED to see the logical models, their assertions, and their physical links.
  • Files are loaded leniently by default (strict_validation: false) so that contracts with extra or non-conformant fields are accepted with warnings. Spec-valid fields the source does not map (SLA, support, pricing, relationships, …) are summarized once per file as an info; genuinely unknown fields warn individually. Set strict_validation: true to fail on JSON Schema violations — recommended for multi-tenant or untrusted directories.
  • Stale-metadata removal uses standard stateful ingestion. Enable it via the stateful_ingestion block (server-side checkpoints; requires a DataHub graph). When you remove a schema[] entry from a contract file, the corresponding logical odcs Dataset and Assertion URNs are marked removed on the next run. Physical datasets and their logicalParent links are never marked removed — those are owned by their platform-of-record source. The fail_safe_threshold guard (default 75%) blocks mass deletions caused by config or naming changes.
  • Symlinks are not followed by default (follow_symlinks: false). If your directory layout organises contracts using symlinks, set follow_symlinks: true deliberately. The default is conservative because following symlinks in a shared directory can disclose files outside the configured root.
  • Files larger than max_input_file_bytes (default 5 MB) are skipped with a warning before parsing.

Install the Plugin​

pip install 'acryl-datahub[odcs]'

Starter Recipe​

Check out the following recipe to get started with ingestion! See below for full configuration options.

For general pointers on writing and running a recipe, see our main recipe guide.

source:
  type: odcs
  config:
    # Path to a single ODCS YAML file, a directory of contracts, or a glob.
    # May also be a list of any of the above. Each file may describe multiple
    # tables (one `schema[]` entry per table) — every entry becomes its own
    # logical `odcs` Dataset carrying the contract's schema and assertions.
    path: ./contracts/

    # Physical binding (the logicalParent link) is derived automatically from
    # each contract's spec-required `servers[].type` — no mapping is needed for
    # postgres / mysql / snowflake / bigquery / redshift / databricks /
    # sqlserver / trino servers. Optional overrides refine env or
    # platform_instance for a named `servers[].server` value:
    # server_overrides:
    #   - server: prod-snowflake
    #     env: PROD
    #     platform_instance: prod

    # Explicit per-table physical URNs (contract id -> schema entry name -> URN).
    # An empty string deliberately leaves that entry unbound; absent names fall
    # back to server-based derivation.
    # physical_urn_overrides:
    #   "orders.v1":
    #     orders: "urn:li:dataset:(urn:li:dataPlatform:snowflake,prod.sales.orders,PROD)"
    #     legacy_orders: ""

    # Standard stateful ingestion: removing a schema[] entry from a contract
    # marks its logical dataset + assertions removed on the next run. Physical
    # datasets are never touched.
    # stateful_ingestion:
    #   enabled: true
    #   remove_stale_metadata: true

    # Optional knobs — uncomment to override defaults.
    #
    # verify_physical_urns_exist: true   # with a DataHub graph: skip logicalParent links whose derived URN doesn't exist
    # lowercase_physical_urns: true      # lowercase composed physical names for platforms that lowercase by default (snowflake)
    # emit_assertions: true              # assertions from quality[] rules, attached to the logical dataset
    # emit_schema_assertion: true        # one DATA_SCHEMA assertion per schema[] entry pinning the contract's schema
    # schema_assertion_compatibility: SUPERSET   # or EXACT_MATCH / SUBSET
    # emit_logical_parent: true          # the physical->logical PhysicalInstanceOf link
    # tag_prefix: "odcs."                # prepend to every emitted tag
    # strict_validation: false           # set true to reject contracts that fail JSON Schema validation
    #
    # Owner identity normalization: ODCS `team[].username` is a username OR an
    # email. Pick at most one of these to match how your identity source keys
    # users. Unresolved owners are reported (report.owners_unresolved) but
    # still emitted — they resolve once the user is provisioned.
    # strip_owner_email_domain: false    # alice@acme.com -> corpuser alice
    # owner_email_domain: acme.com       # alice -> corpuser alice@acme.com
    # odcs_versions: ["3.0.0", "3.0.1", "3.0.2", "3.1.0"]
    # logical_dataset_name_template: "{contract_id}.{schema_name}"  # logical odcs dataset URN name
    # follow_symlinks: false             # follow symlinks when scanning a directory
    # file_extensions: [".yaml", ".yml", ".odcs.yaml", ".odcs.yml"]
    # max_input_file_bytes: 5242880      # 5 MB; larger files are skipped with a warning
    #
    # Replicate contract-level ownership/tags to every logical dataset on each
    # ingest. Set false to emit these aspects only at first sight, which
    # preserves manual UI edits but means contract changes don't propagate.
    # replicate_contract_metadata: true

# Default sink is datahub-rest and doesn't need to be configured.
# See https://docs.datahub.com/docs/metadata-ingestion/sink_docs/datahub for customization.

Config Details​

Note that a . is used to denote nested fields in the YAML recipe.

FieldDescription
path ✅
One of string, array
Path to an ODCS YAML file, a directory containing ODCS YAML files, or a glob pattern. May also be a list of any of the above.
path.string
string
emit_assertions
boolean
Whether to emit Assertion entities derived from the ODCS quality[] rules. Assertions target the logical odcs dataset and are emitted whether or not a physical binding resolves.
Default: True
emit_logical_parent
boolean
Whether to emit a logicalParent link from each resolved physical dataset to its logical ODCS dataset (the PhysicalInstanceOf relationship). Disable to keep ODCS from writing any aspect onto physical datasets.
Default: True
emit_schema_assertion
boolean
Whether to emit one DATA_SCHEMA assertion per schema[] entry, pinning the contract's declared schema on the logical dataset so schema drift is evaluable as a contract violation.
Default: True
follow_symlinks
boolean
If true, follow symlinks when discovering ODCS files. Off by default to prevent disclosure via symlink escape in shared/multi-tenant directories. When enabled, the source still requires that resolved targets stay within the configured root.
Default: False
logical_dataset_name_template
string
Template for the logical odcs dataset name (the URN name segment). Available placeholders: {contract_id}, {schema_name}, {contract_version}.
Default: {contract_id}.{schema_name}
lowercase_physical_urns
boolean
Lowercase composed physical dataset names for platforms whose DataHub connectors lowercase URNs by default (currently snowflake). Set False if your snowflake ingestion runs with convert_urns_to_lowercase disabled. Logical odcs dataset URNs always preserve the contract's casing.
Default: True
max_input_file_bytes
integer
Maximum size of an ODCS YAML file to load. Files exceeding this are skipped with a warning. Defaults to 5 MB. This guards the YAML parser from unbounded inputs.
Default: 5242880
owner_email_domain
One of string, null
When set, bare (non-email) team usernames get @<domain> appended, so alice maps to corpuser alice@acme.com — use this when your identity source ingests users by email. Mutually exclusive with strip_owner_email_domain.
Default: None
physical_urn_overrides
map(str,map)
replicate_contract_metadata
boolean
If True (default), contract-level Ownership and GlobalTags are written to the logical dataset on every ingest run. Set False to skip emitting contract-level Ownership and GlobalTags so manual UI edits to those aspects survive subsequent ingest runs (the contract is then a one-time enricher rather than a source of truth).
Default: True
schema_assertion_compatibility
string
Compatibility mode for the DATA_SCHEMA assertion: SUPERSET (an instance must contain at least the contract's fields; extras allowed), EXACT_MATCH, or SUBSET.
Default: SUPERSET
strict_validation
boolean
If True, ODCS files failing JSON-Schema validation are skipped (logged as warnings). If False (default), schema violations are reported as warnings; the contract is still ingested. Default False matches real-world ODCS files that use deprecated forms (e.g. top-level quality[]); set True to fail-fast on schema violations.
Default: False
strip_owner_email_domain
boolean
ODCS team usernames may be emails. When true, the domain is stripped so alice@acme.com maps to corpuser alice — use this when your identity source (Okta / Azure AD / …) ingests users by bare username. Mutually exclusive with owner_email_domain.
Default: False
tag_prefix
One of string, null
Optional prefix prepended to every tag emitted from ODCS tags and property-level tags (e.g. odcs.). Useful for distinguishing ODCS-sourced tags.
Default: None
verify_physical_urns_exist
boolean
When a DataHub graph is available (datahub-rest sink), verify each derived physical dataset URN exists before emitting its logicalParent link; URNs not found are left unbound with a warning instead of creating stub datasets. With no graph (file sink), emission proceeds without verification. Set False to emit links optimistically for tables that do not exist yet.
Default: True
env
string
The environment that all assets produced by this connector belong to
Default: PROD
file_extensions
array
File extensions considered ODCS YAML files when scanning a directory.
file_extensions.string
string
odcs_versions
array
List of supported ODCS apiVersion values. Contracts with apiVersion outside this list are skipped with a warning.
odcs_versions.string
string
server_overrides
array
Optional per-server overrides for physical binding. The physical platform is derived from the contract's spec-required servers[].type; an override refines env / platform_instance / platform for a named servers[].server value. Binding only affects the logicalParent link — assertions always attach to the logical dataset.
server_overrides.ServerMapping
ServerMapping
Per-server overrides for physical dataset binding.

The physical platform is normally derived from the spec-required
servers[].type; an override entry refines that derivation for a named
servers[].server — most commonly to set env or platform_instance,
or to force a different platform id.
server_overrides.ServerMapping.server ❓
string
The value of servers[].server in the ODCS contract. If match_any is True, this is treated as a wildcard match for any server.
server_overrides.ServerMapping.match_any
boolean
If True, this mapping matches any servers[].server not matched by a more specific mapping. Use a single catch-all mapping to apply one override to all contracts.
Default: False
server_overrides.ServerMapping.platform
One of string, null
Optional DataHub data platform identifier (e.g. postgres, snowflake). When unset, the platform is derived from the server's type.
Default: None
server_overrides.ServerMapping.platform_instance
One of string, null
Optional DataHub platform instance, used when the same platform is deployed multiple times (e.g. multiple Snowflake accounts).
Default: None
server_overrides.ServerMapping.env
string
The environment / fabric type for the produced physical dataset URNs (e.g. PROD, DEV).
Default: PROD
stateful_ingestion
One of StatefulStaleMetadataRemovalConfig, null
Stateful Ingestion Config
Default: None
stateful_ingestion.enabled
boolean
Whether or not to enable stateful ingest. Default: True if a pipeline_name is set and either a datahub-rest sink or datahub_api is specified, otherwise False
Default: False
stateful_ingestion.fail_safe_threshold
number
Prevents large amount of soft deletes & the state from committing from accidental changes to the source configuration if the relative change percent in entities compared to the previous state is above the 'fail_safe_threshold'.
Default: 75.0
stateful_ingestion.remove_stale_metadata
boolean
Soft-deletes the entities present in the last successful run but missing in the current run with stateful_ingestion enabled.
Default: True

Capabilities​

Use the Important Capabilities table above as the source of truth for which DataHub features this source emits. The notes below cover ODCS-specific behavior that the table does not capture.

Quality rule mapping​

Each entry in a schema[] (table-level) or properties[] (column-level) quality[] array becomes a DataHub assertion attached to the logical odcs dataset, emitted whether or not a physical binding resolves. The library vocabulary is spec-exact and version-dependent:

ODCS ruleDataHub aspect
v3.1 metric: nullValues with mustBe: 0FieldAssertionInfo + FieldValuesAssertion (NOT_NULL)
v3.1 metric: nullValues with another thresholdFieldAssertionInfo + FieldMetricAssertion (NULL_COUNT / NULL_PERCENTAGE)
v3.1 metric: duplicateValues / v3.0 rule: duplicateCount with mustBe: 0FieldAssertionInfo + FieldMetricAssertion (UNIQUE_PERCENTAGE == 100)
v3.1 metric: invalidValues (arguments.validValues) / v3.0 rule: validValuesFieldAssertionInfo + FieldValuesAssertion (IN)
v3.1 metric: invalidValues with arguments.patternFieldAssertionInfo + FieldValuesAssertion (REGEX_MATCH)
metric: rowCount (v3.0 rule: rowCount)VolumeAssertionInfo (unit: rows only)
type: sql with a query and a mappable thresholdSqlAssertionInfo
type: custom (engine + implementation)CustomAssertionInfo (type = the engine, logic = the implementation)
type: textCustomAssertionInfo (logic = the description)
Anything else with contentCustomAssertionInfo (original rule intent preserved as logic)
Anything with no operator and no bodySkipped with a warning

The v3.1 library key is metric; rule is accepted as the v3.0 canonical key and as the deprecated v3.1 alias (v3.1 documents using rule get an informational notice). Every assertion carries ODCS provenance in customProperties (odcs.id, odcs.rule.id, odcs.rule.metric, odcs.rule.unit, serialized odcs.rule.arguments, dimension / severity / businessImpact when present), a dataPlatformInstance aspect attributing it to the odcs platform, and an externalUrl from the rule's authoritativeDefinitions when provided. Assertion URNs are seeded from the spec's quality.id when present, so renames and reordering do not churn identities.

Routing is exact; nothing is approximated

Tolerances map to typed assertions only when they are exactly representable:

  • No threshold at all (the v3.0 validValues form) and mustBe: 0 both mean "no failing rows tolerated".
  • An integral mustBeLessOrEqualTo maps to a fail threshold in rows (unit: rows) or percent (unit: percent).
  • Everything else — strict less-than tolerances, non-integral percents, mustNotBeBetween, duplicate-count tolerances, percent-based rowCount, multi-column duplicateValues (arguments.properties), missingValues — is preserved as a CustomAssertionInfo whose logic carries a stable rendition of the original rule (metric, arguments, thresholds, unit) rather than being approximated into a typed shape.
  • Rules routed to custom are listed in report.rules_routed_to_custom; rules with no modelable content at all are skipped and listed in report.rules_skipped_no_threshold.

Schema-compliance assertion​

For every schema[] entry that declares properties, the source emits one DATA_SCHEMA assertion (SchemaAssertionInfo) on the logical dataset carrying the contract's declared schema. This makes schema drift an evaluable contract violation rather than an implied one. schema_assertion_compatibility controls the mode: SUPERSET (default — an instance must contain at least the contract's fields, extras allowed), EXACT_MATCH, or SUBSET. Disable with emit_schema_assertion: false.

Binding exists to link physical datasets to their logical model — assertions never depend on it. Resolution per schema[] entry, in priority order:

  1. physical_urn_overrides[<contract id>][<schema entry name>] — an explicit URN, or an empty string to deliberately leave the entry unbound. Entries absent from the map fall back to derivation; keys that match no schema entry warn.
  2. The contract's first mappable servers[] entry. The platform comes from the spec-required servers[].type (or a matching server_overrides entry), and the table name (physicalName, falling back to name) is qualified with the server's own fields per the platform's URN convention:
Server typeDataHub platformPhysical name
postgrespostgresdatabase.schema.table
redshiftredshiftdatabase.schema.table
sqlservermssqldatabase.schema.table
snowflakesnowflakedatabase.schema.table (lowercased by default; lowercase_physical_urns: false opts out)
bigquerybigqueryproject.dataset.table
databricksdatabrickscatalog.schema.table
trinotrinocatalog.schema.table
mysqlmysqldatabase.table
oracleoraclenot composable — supply a dotted physicalName or an explicit override
anything else—unbound (logical dataset and assertions unaffected)

A physicalName that already contains a dot is used verbatim (assumed pre-qualified) and counted in report.physical_names_passthrough. Missing server fields leave the entry unbound with an actionable reason — the source never guesses a schema name. When a DataHub graph is available and verify_physical_urns_exist is on (default), derived URNs that do not exist in DataHub are left unbound with a warning instead of creating stub datasets. Two schema entries binding the same physical dataset warn: logicalParent is single-valued, so the last writer wins.

Limitations​

  • ODCS v3.0 and v3.1 only. Contracts reporting v2.x in apiVersion are skipped with a warning.
  • Logical Models are in private beta and render in the UI only when LOGICAL_MODELS_ENABLED is enabled (off by default). The logical datasets and their assertions are ingested while the flag is off — they just aren't displayed.
  • Nested-column assertions and field-path resolution. Field-scoped assertions reference the column by its dotted property path (address.city), matching the logical dataset's own schemaMetadata. Propagation onto physical datasets whose connectors encode nested struct paths differently is subject to the platform propagation mechanism.
  • Property-level unique: true / required: true flags are not emitted as assertions — they map into schemaMetadata (nullability, keys) and are enforced via the schema-compliance assertion; express uniqueness checks as quality[] rules (metric: duplicateValues with mustBe: 0) to get a typed field assertion.
  • File loading is capped at 5 MB by default (max_input_file_bytes). Larger YAML files are skipped with a warning before parsing.
  • Out of scope: SLA (slaProperties), support, price, v3.1 relationships (foreign keys), customProperties → DataHub customProperties (only the odcs.* provenance subset is emitted), classification → GlossaryTerm linking, schemaField-level logicalParent column links, and ODCS export. Spec-valid-but-unmapped fields are reported once per file via report.spec_fields_ignored. These may land in a follow-up.
  • Contract metadata replication: By default, contract-level ownership and tags are written to every logical dataset on each run. If you edit these aspects in the DataHub UI, they will be overwritten on the next ingest. Set replicate_contract_metadata: false to disable replication (useful for one-time enrichment workflows).
  • Mixed-platform servers[]: binding uses the first server whose type maps to a platform; contracts that reference multiple platforms bind to that one only. Use physical_urn_overrides for per-table control.

Troubleshooting​

Where do the emitted assertions appear?​

On the logical odcs dataset — open it in the UI (requires LOGICAL_MODELS_ENABLED) and look under its Quality / Assertions tab. Assertions are not written to physical datasets by this source; DataHub propagates expectations to physical instances via the PhysicalInstanceOf relationship.

Check report.unmappable_servers and the per-entry info messages: the contract may declare no servers[], use a server type with no platform mapping (e.g. kafka, s3), or be missing the fields needed to qualify a table name (e.g. a postgres server without schema). With a DataHub graph attached, report.physical_urns_unverified counts derived URNs that were skipped because they do not exist in DataHub yet — ingest the physical platform first, or set verify_physical_urns_exist: false to link optimistically.

My logical odcs dataset doesn't appear in the UI​

Logical Models are in private beta and require the LOGICAL_MODELS_ENABLED feature flag (off by default). Enable it to view logical datasets, their assertions, and their logicalParent links. The metadata is still ingested while the flag is off — it just isn't displayed. The flag is set on the GMS service (for self-hosted OSS, the LOGICAL_MODELS_ENABLED environment variable on the datahub-gms container; on DataHub Cloud, ask your administrator to enable it).

My missingValues/mustNotBeBetween/vendor rule shows up as a Custom Assertion​

Expected. See Quality rule mapping for the exact typed-assertion allowlist and the threshold-representability rules. Anything outside them is preserved verbatim as CustomAssertionInfo rather than approximated.

I removed a table from my ODCS file but the physical dataset is still in DataHub​

That's expected. ODCS owns the logical odcs Dataset and the Assertions it emitted; the physical Dataset belongs to its platform-of-record source (postgres / snowflake / …), and the logicalParent link is a non-destructive enrichment. With stateful_ingestion enabled, removing the schema[] entry marks the logical dataset and its assertions removed on the next ODCS ingest — never the physical dataset.

My contract owner shows up as an unresolved user​

ODCS team[].username is a username or an email, and DataHub resolves the owner against whatever identifier your identity source (Okta / Azure AD / SCIM / …) ingests users under. If the two disagree — the contract says alice@acme.com but DataHub knows the user as alice, or vice versa — the ownership reference dangles. Set exactly one of strip_owner_email_domain: true (emails → local part) or owner_email_domain: acme.com (bare usernames → emails) to normalize contract identifiers to your convention; explicit urn:li:corpuser: / urn:li:corpGroup: values pass through untouched. When ingesting through a DataHub graph, owners that don't resolve are flagged in the run report (report.owners_unresolved, one warning per unique owner) — the reference is still emitted and becomes functional as soon as the user is provisioned.

I edited owners on a logical dataset and the next ingest reverted them​

By default, contract-level ownership replicates on every ingest, so any UI edits are overwritten. Set replicate_contract_metadata: false in the source config to switch to first-sight-only emission, which preserves UI edits after the initial ingest. The same setting also governs top-level tags.

For the canonical ODCS spec and JSON Schemas, see the open-data-contract-standard repository. For an audience-friendly entry point and adopter list, see bitol.io.

Code Coordinates​

  • Class Name: datahub.ingestion.source.odcs.odcs_source.ODCSSource
  • Browse on GitHub
Questions?

If you've got any questions on configuring ingestion for Open Data Contract Standard, feel free to ping us on our Slack.

💡 Contributing to this documentation

This page is auto-generated from the underlying source code. To make changes, please edit the relevant source files in the metadata-ingestion directory.

Tip: For quick typo fixes or documentation updates, you can click the ✏️ Edit icon directly in the GitHub UI to open a Pull Request. For larger changes and PR naming conventions, please refer to our Contributing Guide.