[Bug][iceberg] REST commit fails on format-version 3: Cannot parse missing long: next-row-id

[vbabenkoru]

Search before asking

• I searched in the issues and found nothing similar.

Paimon version

2.0.0-rc4

Compute Engine

Flink 2.2.0

Minimal reproduce step

CREATE TABLE t (
  project_id BIGINT,
  email STRING,
  ...
  PRIMARY KEY (project_id, email) NOT ENFORCED
) PARTITIONED BY (project_id) WITH (
  'bucket' = '16',
  'changelog-producer' = 'lookup',
  'metadata.iceberg.storage' = 'rest-catalog',
  'metadata.iceberg.format-version' = '3',
  -- or enable deletion-vectors (docs: DV requires format-version 3)
  'metadata.iceberg.rest.uri' = 'https://glue.<region>.amazonaws.com/iceberg',
  'metadata.iceberg.rest.warehouse' = '<account-id>',
  'metadata.iceberg.rest.io-impl' = 'org.apache.iceberg.aws.s3.S3FileIO',
  'metadata.iceberg.rest.rest.sigv4-enabled' = 'true',
  'metadata.iceberg.rest.rest.signing-name' = 'glue',
  'metadata.iceberg.rest.rest.region' = '<region>',
  'metadata.iceberg.rest.client.credentials-provider' =
    'software.amazon.awssdk.auth.credentials.WebIdentityTokenFileCredentialsProvider'
);

-- insert / Flink sink until a Paimon snapshot commits

Inspect generated metadata (example): format-version is 3, but top-level next-row-id is absent. Snapshots may carry first-row-id / added-rows, but table-level next-row-id is missing — Iceberg 1.11 requires it for v3.

What doesn't meet your expectations?

The Flink job fails on a commit with error Cannot parse missing long: next-row-id

Anything else?

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!

[vbabenkoru]

Iceberg v3 row lineage requires table metadata next-row-id (see Iceberg TableMetadataParser). Paimon’s IcebergMetadata.toJson() emits v3 without that field, so TableMetadataParser.fromJson fails before any REST createTable/updateTable.