## July 27, 2026

*** ** * ** ***

Announcement Google Distributed Cloud (GDC) air-gapped 1.16.2 is available.  
See the [product overview](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/overview.md.txt) to learn about the features of Distributed Cloud.

*** ** * ** ***

Feature The following new features are available:

**Backup and restore**

- Added the `maxConcurrency` attribute in the `spec` field of the `ClusterBackupPlan` and `VirtualmachineBackupPlan` custom resources. For more information, see [Plan backups](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform-application/pa-ao-operations/cluster-backup/plan-backups.md.txt) and [Create a scheduled backup plan](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform-application/pa-ao-operations/vm-backup/backup-plans/create-scheduled-backup-plan.md.txt).

**Cluster management**

- Added support for [creating](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/create-standard-cluster.md.txt#console), [deleting](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/delete-standard-cluster.md.txt#console), [editing](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/manage-node-pools.md.txt#console), and [viewing](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform-application/pa-ao-operations/cluster.md.txt#console_2) standard Kubernetes clusters in the GDC console.

**Networking**

- Added support for
  [configuring explicit deny project network policies](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/pnp/deny-traffic.md.txt)
  as a [Preview](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/resources/feature-stages.md.txt#preview) feature to
  restrict connectivity between workloads and services in different projects
  or CIDR ranges.

- Added support for targeting specific cluster nodes by IP address or CIDR
  range using the `ipBlock` field in
  [Kubernetes network policies for standard clusters](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/k8s-network-policy/standard-cluster-traffic.md.txt#create-node-policy)
  and the `ipBlocks` field in
  [project network policies](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/pnp/pnp-overview.md.txt#node-selection-considerations)
  as a [Preview](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/resources/feature-stages.md.txt#preview) feature.

**Virtual machines**

- Added a `taints` patch to the `OrganizationZonalConfig` resource for high performance computing (HPC) optimizations.

- Added support for [NVIDIA GPU driver installation on Windows VMs](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/application/ao-user/vms/manage-the-os/install-nvidia-drivers-win.md.txt).

*** ** * ** ***

Issue The following issues are identified:

**Cluster management**

- Kubernetes cluster deletion stops responding because a LUKS secret remains stuck in the terminating namespace.

**Harbor**

- The `gdcloud harbor backup-plans` command group is missing the `update` and `delete` commands, returning a `command not found` error when run.

**Infrastructure as code**

- GitLab CI presubmit linters in the `gdch/iac` repository validate unmodified
  files across the entire codebase, which can block urgent merge requests due
  to pre-existing YAML lint errors.

- There are reconciliation errors during an upgrade from 1.15.2 or
  later.

**Inventory**

- Incorrect data plane network bond generated on `Cell`, `Server`, and `BaremetalHostNetwork` custom resources for B300 (`d4-highgpu2-256-gdc-metal`) servers when `fe-s6p1` and `fe-s6p2` aren't ordered first in `devices.csv`.

**Logging**

- Linux audit logs collected from node-local audit sources lack the `host` field in log output.

**Monitoring**

- Cortex `VirtualService` and `IstioAuthorizationResource` resources, and Grafana `IstioAuthorizationResource` resources, remain in the cluster after a project is deleted.

- Certificate secrets are leaked when deleting a `MonitoringTarget`.

- The Grafana dashboard link from a standard cluster details page in the GDC console incorrectly redirects you to the `platform-obs` dashboard instead of loading data for the selected project.

- You can safely ignore or disable certain high-frequency alerts.

**Networking**

- During an organization upgrade to 1.16.2 hotfix 2 or later on nodes running Rocky Linux 9, the upgrade stops at step `infra-cp.HostNetwork` because the Host Networking component attempts to install Rocky Linux 8 RPM packages; hardware IPsec offloading is unavailable on Rocky Linux 9 nodes.

- In Operations Infrastructure Rack v2 (OIRv2) universes, the management switch doesn't
  advertise the out-of-band management CIDR block to upstream Layer 3 switches
  over Border Gateway Protocol (BGP).

- Custom BFD timer parameters configured on `InterconnectAttachment` custom resources are not propagated to border leaf switch interfaces.

- Hairpin BGP sessions with unassigned VLAN IDs cause physical network upgrade checks and switch reconciliation to fail during upgrades.

- The `hnet_network_interface_down` alert might trigger false-positive warnings
  for physical `eth*` interfaces that are operationally down but unused by
  Host Networking.

**Platform authentication**

- Authentication to tenant organizations fails with a database connection reset error in `ais-storage` pods after upgrading to 1.16.2.

**Security Information and Event Management (SIEM)**

- Subcomponent `siem-aas-global` fails to reconcile in the org management plane.

**Storage**

- After upgrading to 1.16.2 hotfix 3, the `file-user-journey` subcomponent
  fails reconciliation with a missing `FileUserJourneyFramework` feature
  error.

- An issue causes the object storage upgrade request to enter a
  `QualificationFailed` state during upgrades to 1.16.2.

- Default keep lists filter out required metrics from Grafana.

- Dual-zone buckets can't be deleted from the console.

**Ticketing system**

- The Support portal displays browser TLS certificate warnings after upgrading to 1.16.2.

**Upgrade**

- The `sar` preflight upgrade check fails if `gdcloud system container-registry load-oci` is run with `--trigger-signature-verification=true`, because both the command and the organization upgrade preflight check create `artifact-signature-verification` jobs in the `package-validation-system` namespace.

- While upgrading to 1.16.2, the `atat-invoice-export` and `atat-portfolio` subcomponents in the `root` namespace might enter a `ReconciliationError` state.

- The `sar` preflight and postflight upgrade checks fail when you upgrade from
  a hotfix release that doesn't include critical images.

**Vertex AI**

- Vertex AI Service Usage dashboards display a `No data` state because the PromQL queries contain restrictive filters for NVIDIA resources and Workbench labels that might not be present in the environment.

**Virtual machines**

- VMs in tenant organizations that use the package repository servers of the
  `gdchservices` organization can't install or update packages from
  `pkg.org.internal` and receive `503 Service Unavailable` errors, because the
  `pkg-infra-cert` certificate is generated without the DNS suffix.

- The **API** example for attaching a disk to a virtual machine (VM) indents
  `boot` and `autoDelete` under `virtualMachineDiskRef`, causing the
  `kubectl edit` command to fail.

**Vulnerability management**

- During an upgrade from 1.16.1 to 1.16.2, the `lcm.private.gdc.goog/opsservices`
  label is removed from OIRv2 `opsservices` APIs.

- Tenable Nessus offline plugin updates fail due to a missing detached
  signature file.

*** ** * ** ***

Fixed The following issues are fixed:

**Documentation**

- Stale documentation shipped with the product release.

**Cloud DNS**

- Fixed an issue where `oclcm-configrunner` reported a benign `cannot match trustStoreGlobalRootExtResourceVersion` monitoring parameter error for the `dns-customer-resolver` subcomponent.

**GDC console**

- Documentation hosted in the GDC console shows a 404 error.

**gdcloud CLI**

- Running any gdcloud CLI command prints an uninitialized feature flag warning.

**Logging**

- Fixed an issue where alert `AL-E0935` (`log_query_api_list_logs_error_rate_high`) fired when log queries exceeded Loki's 30-day time range limit. Client input errors are now validated upfront and excluded from error rate alerts.

**Networking**

- Enabling Cloud NAT on a VM prevents SSH connections using the GDC console UI.

**Network Time Protocol (NTP)**

- When deploying OIRv2 management switches, NTP configuration and encryption might not apply to the running configuration.

**Servers**

- Bootstrapper installation fails during file system configuration with a `matched no disk` error.

**Storage**

- The ONTAP S3 event audit logging is disabled.

**Ticketing system**

- The `ts-app-server-1` pod restarts periodically.

**Virtual machines**

- VM creation with types `a3-ultragpu-8g` and `a4-ultragpu-8g` on a newly
  reprovisioned server fails with errors.

- `gpu-controller-controller-manager pod` enters an `ImagePullBackOff` state during upgrade to release 1.16.1.

**Vulnerability management**:

- Vulnerability management subcomponents and system `UserAccount` SSH key generation might get stuck due to a container runtime deadlock and Helm parameter cache poisoning.

*** ** * ** ***

Change The following changes are identified:

**Version updates**:

- The Google Distributed Cloud for bare metal version is updated to 1.32.1300-gke.46 to apply the latest security patches and important updates.

*** ** * ** ***

Deprecated The following features are deprecated:

**Database services**

- The Database Service for PostgreSQL is now powered by AlloyDB Omni. Database Service for AlloyDB Omni will no longer be available as a standalone managed service. Instead, customers looking to use AlloyDB Omni as a managed service should transition to the Database Service for PostgreSQL, which is powered by AlloyDB Omni. Additionally, Database Service for Oracle, a managed offering, is deprecated but the option to self-manage Oracle deployments remains fully supported. For more information on self-managing Oracle databases, refer to the [solution guides](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/db-self-managed-oracle-ref-arc). To support your planning, starting June 29, 2026 and ending on June 29, 2027, we are providing a 12-month discontinuation period for Database Service for AlloyDB Omni and Database Service for Oracle.

**Virtual machines**

- Distributed Cloud deprecates enabling external access using `VirtualMachineExternalAccess` during VM creation in the GDC console. To manage external networking, create an External Load Balancer (ELB) for ingress and a Cloud NAT Gateway for egress.