Skip to main content

Control Plane

Recent Control Plane releases — Gateway Controller, Infra, Ilake Envoy, and Customer Stack (AWS, GCP, Azure). Showing the 10 most recent versions.

release-v0.265.1 — 2026-08-20​

General​

  • CDC bootstrap heartbeats are now progress-gated per table, reducing unnecessary heartbeat traffic and providing more accurate per-table bootstrap progress reporting.
  • Metrics extraction now correctly honours configured exclusion filters, preventing excluded metrics from being collected.
  • BigQuery sync jobs now support the bq.sync.table.name configuration option for controlling the target table name.
  • Spark code jobs now enforce a concurrent-run admission limit, preventing duplicate or runaway parallel executions of the same job.
  • Versioned table entries are no longer incorrectly deactivated when a stream capture is stopped, preserving table version history across stream lifecycle events.
  • The startup-script watchdog timeout has been raised from 60 seconds to 120 seconds, reducing spurious watchdog-triggered restarts during slower node initializations.
  • The Spark log storage path is now persisted on the job and job run records and returned by DESCRIBE JOB and DESCRIBE JOB_RUN.
  • Iceberg table records in the OneHouse catalog are now populated with metadata location hints and metastore type, enabling more accurate catalog integrations.
  • Spark jobs now support per-run Hudi configuration overrides, and temporary Hudi config files are automatically cleaned up when a run completes.
  • Spark drivers are now equipped with a heap-guard and the driver liveness probe threshold has been raised to 600 seconds, improving resilience under high memory pressure.
  • User-configured storage base paths are now automatically normalized to the correct scheme, resolving storage access errors caused by mismatched path formats.
  • The control plane now retries fetching database credentials at startup, reducing pod initialization failures caused by transient secret-fetch errors.
  • The OneHouse catalog now supports Iceberg write operations, including table creation and commit, enabling write-path workflows against the native catalog.
  • Clusters can now be created or renamed using the name "Default Managed Cluster", which was previously incorrectly rejected.

release-v0.264.3 — 2026-08-17​

General​

  • Tables in observed lakes can now be unregistered to stop observation without deleting the underlying data.
  • The SQL API now supports ALTER FLOW IN TABLE, allowing a flow's processing cluster to be moved by specifying the target lake, database, and table.
  • Duplicate entries showing both in-progress and completed states for the same commit no longer appear in the Table Service history view.
  • User management list views now immediately reflect the latest changes after an update, eliminating stale reads.
  • Stale access-control grants and groups are now automatically revoked and cleaned up when they are no longer needed.
  • ACL synchronization operations are now retried automatically when contention is detected, reducing catalog sync failures.
  • The catalog server endpoint is now surfaced and available for client connections.
  • Spark catalog jobs now use the unified REST catalog, enabling consistent cross-format table access.
  • Fixed a production issue where AWS S3A-based orchestrator workflows failed due to an AWS SDK version conflict.

release-v0.262.2 — 2026-08-10​

General​

  • A runtime error causing community-edition deployments on AWS to crash with a missing field in the AWS SDK has been fixed.
  • Dataplane cluster configuration responses now include node pod-capacity limits, enabling more accurate node utilization and scheduling decisions.
  • Organizations can now create and manage a OneHouse catalog through the external API, with availability controlled on a per-organization basis.

release-v0.261.1 — 2026-08-04​

General​

  • Cross-account Kinesis ingestion sources now accept a role ARN parameter on CREATE SOURCE, enabling streaming ingestion from Kinesis streams hosted in a different AWS account.
  • The number of tables configurable in a single Debezium ingestion workflow is now capped, preventing oversized configurations that could cause instability.
  • Compute clusters that encounter a cancelled state write during deletion no longer get permanently stuck in the DELETING state.
  • A dependency conflict that caused scheduled jobs to crash at startup has been resolved.

release-v0.260.2 — 2026-07-30​

General​

  • Scheduled jobs no longer crash at startup due to an AWS SDK version conflict.
  • Additional Azure instance types (including Standard_E8pds_v6) are now available for cluster configuration.
  • Spark workloads on single-stack IPv6 clusters now connect reliably to the Kubernetes API.
  • Creating a stream capture is now blocked when a MySQL or PostgreSQL connector is in a degraded state, preventing pipelines from starting in a broken condition.
  • The OneHouse job run ID is now injected into the Spark job runtime environment, making it available for tracing and correlation.
  • The SQL endpoint is automatically refreshed when a stale connection is detected on cluster read, preventing connection failures after endpoint migrations.
  • Bootstrap connector failures no longer incorrectly block new stream capture creation for a source.
  • Unity Catalog integrations now support customer-managed secrets (BYOS) for credential configuration.

release-v0.258.0 — 2026-07-22​

Bug Fixes​

  • Fixed a crash during CDC bootstrap when two or more tables share the same source table, which caused snapshot-status reads to fail with a duplicate-key error.
  • Fixed a CDC bootstrap hang and erroneous PAUSED-to-active resurrection caused by bootstrap-polling gates not counting agent-reported tables.
  • Fixed an issue where the account-level OCU usage page showed no values for projects not linked to a billing account.
  • Account-level OCU usage now correctly reports unlinked projects as zero consumption instead of silently omitting them.
  • Fixed Kafka ingestion into append-only CDC transformer streams, which previously failed due to incorrect payload class handling.
  • Fixed compute cluster stop and delete operations that could fail when a lock provider was not present during validation.
  • Fixed an issue where bulk-starting clusters with attached storage could fail to materialize node pools due to a stale storage configuration.

General​

  • Cross-account Kinesis stream access is now supported in the onboarding CloudFormation template, enabling ingestion from Kinesis streams in external AWS accounts.
  • Account-level OCU consumption can now be queried with per-project filtering, giving account administrators a more granular view of usage across their projects.
  • Spark driver and executor failure exit codes are now tracked via metrics, improving visibility into job failure root causes.
  • Users now receive email and in-app notifications when a just-in-time support access request is created for their environment.

release-v0.257.1 — 2026-07-20​

General​

  • Attached storage configuration is now correctly refreshed when a compute cluster is restarted or stopped and started, preventing node pools from being left in an inconsistent state.
  • Email and in-app notifications are now sent when a JIT support access request is created.
  • The billing checkout flow now handles redirect completion more reliably, reducing the likelihood of stuck checkout sessions.
  • CloudFormation onboarding trust policies are now consistent with the Terraform-based onboarding templates.
  • S3 SNS notification conflicts are now detected at source creation time, surfacing configuration errors earlier in the ingestion setup flow.
  • The core IAM role in CloudFormation onboarding now has read access to BYOS telemetry secrets, fixing a permissions gap for BYOS-configured environments.
  • Job and run URLs returned by the DESCRIBE JOB API are now correct.

release-v0.256.1 — 2026-07-09​

General​

  • Job and run links in DESCRIBE JOB output now resolve correctly.
  • Completed Merge-on-Read compaction operations now appear correctly in the table commit history instead of showing as perpetually in progress.
  • Free-tier plan downgrades are now limited to twice per calendar month.
  • Table metrics now load faster by eliminating blocking waits in the metrics handler.
  • The default payment method is now reliably pinned whenever a subscription transitions to active, preventing checkout failures.
  • Log forwarding to Amazon S3 is now supported and automatically applied when the observability exporter is configured.
  • Configuring a DynamoDB lock provider with leading or trailing spaces in the table name now returns a clear validation error instead of an unexpected failure.

release-v0.255.0 — 2026-07-09​

General​

  • DESCRIBE JOB and DESCRIBE JOB_RUN now return the job's unique identifier and a direct UI link to the job run.

release-v0.254.1 — 2026-07-09​

General​

  • Grafana organization credentials are now only accessible when customer monitoring is explicitly enabled.
  • Compaction now defaults to an unbounded I/O budget, preventing log-file-size-based compaction from being artificially throttled.
  • Scheduled jobs now automatically retry fetching authentication tokens on transient failures, improving reliability.
  • Connector retry logic no longer unnecessarily restarts connectors that are already healthy.
  • Grafana organization usernames can now be configured directly through the UI instead of being hardcoded.
  • Fixed a bug where per-organization configuration lookups could silently use stale settings from a different organization.
  • Deleting a table now correctly invalidates its cached metadata, preventing stale data from being served after deletion.
  • The CDC source retry button now remains disabled until the next health check completes, preventing premature retries.
  • CDC source health status is now aggregated per source, and all connectors belonging to a failed source are restarted together.