## Hotfix upgrade instructions for version 1.15.3

You must account for the following requirements to ensure a successful upgrade
to GDC 1.15.3:

- For new bootstraps, you must bootstrap the root organization directly to the
  1.15.3-gdch.3086 binary.

  > [!IMPORTANT]
  > **Important:** Load all 1.15.3 hotfixes *before* creating any new organizations.

- To prevent a critical operating system (OS) downgrade when upgrading your
  environment from version 1.14.12, you must complete the following:

  1. Load all available hotfixes and version 1.15.3-gdch.3086 into the system
     first.

  2. Verify hotfixes are loaded prior to starting the upgrade to ensure the
     correct OS version is used.

## Hotfix 8

Fixed The 1.15.3-gdch.3086-8 hotfix is available. This hotfix fixes the following issue:

**Identity and Access Management**

- Removed overprivileged write permissions for custom roles (`iamroles`, `standardclusterroles`, and `projectroles`) from the global and zonal `project-iam-admin` predefined roles, restricting custom role creation and modification capabilities to prevent unauthorized privilege escalation for application operators.

## Hotfix 7

Fixed The 1.15.3-gdch.3086-7 hotfix is available. This hotfix fixes the following issue:

**Virtual machines**

- Added [Secure Boot certificate renewal](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/application/ao-user/vms/manage-vms/secure-vms/secure-boot-cert-update.md.txt) support for VMs with expiring certificates.

## Hotfix 6

Fixed The 1.15.3-gdch.3086-6 hotfix is available. This hotfix fixes the following issue:

**Networking**:

- Fixes an issue where Internet Protocol Security didn't work because there were too many requests to the server.

## Hotfix 5

Fixed The 1.15.3-gdch.3086-5 hotfix is available. This hotfix fixes the following issues:

**Operating system**:

- Added SLO dashboards for operating systems in the root admin cluster and
  management API server.

- Fixes the missing delete permission for the `OSPolicy` controller.

## Hotfix 4

Fixed The 1.15.3-gdch.3086-4 hotfix is available. This hotfix fixes the following issues:

**Backup**:

- Prevents backup plans from running overlapping backups to conserve system
  resources, especially against aggressive schedules.

- Fixes an issue where the `VirtualMachineRestore` object couldn't be deleted
  properly.

- Fixes an issue where an imported backup without a creation time couldn't be
  deleted.

**Billing**:

- Introduces a new billing SKU required for Gemini 2.5 Flash.

**Cluster management**:

- Fixes an issue where the Kubernetes cluster is not removed after deletion.

**Database Service**:

- Improves the reliability of creating and deleting databases and backups.

**Cloud DNS**:

- Fixes the issue where wildcard DNS records aren't created at the zone apex
  level.

- Fixes the issue where it takes over 25 minutes to create a managed DNS zone.

**Health**:

- Prevents alerts for SLOs marked as experimental by submitting an Infrastructure as Code (IaC) change.

**Inventory management**:

- Fixes the issue where the `gdcloud system assets add` command failed to generate `splitInterface` configurations for breakout cables during dynamic expansion.

**Lifecycle management**:

- Fixes an issue that prevented successful upgrades for subcomponents that used the `lcm.private.gdc.goog/paused-remote: "true"` annotation.

**Logging**:

- Fixes the egress label of the org infrastructure cluster for external security information and event management.

**Networking**:

- Fixes an issue where a cluster gets stuck in a deleting state.

- Adds new networking dashboards for runtime metrics.

**Platform authentication**:

- Adds additional probers to the system.

**Security**:

- Fixes the SSH machine certificate generation process so it's an atomic
  operation.

- Fixes an issue during the Nessus activation process for new installations.

**Storage**:

- Fixes the persistence of dual-zone buckets after deletion in the
  GDC console.

- Fixes the storage bucket versioning errors in the GDC console.

- Fixes the issue where security key materials could be mismatched across zones
  in a dual-zone bucket.

- Fixes the issue where the StorageGRID load balancer endpoint server
  certificates don't automatically rotate.

**Upgrade**:

- Fixes an issue that causes the `siem-cluster` subcomponent to become unresponsive.

**Virtual machines**:

- Fixes data population issues in the following IO dashboards:

  - VMM-R0006: VM SSH connection problems
  - VMM Control Plane: pods (expected versus observed)
  - VMM GPU Control Plane: pods (expected versus observed)
- Fixes an erroneous alert that occurs when creating a virtual machine disk.

## Hotfix 3

Fixed The 1.15.3-gdch.3086-3 hotfix is available. This hotfix fixes the following issues:

**Security**:

- Updates Transport Layer Security (TLS) standards to restrict and enforce approved cipher suites (AES-256).

## Hotfix 2

Fixed The 1.15.3-gdch.3086-2 hotfix is available. This hotfix fixes the following issues:

**Firewall**:

- The `FirewallNode` resource is in a `NotReady` state after upgrade.

**Networking**:

- The `unet-root-admin-cm` job fails due to a missing organization.

**Node OS**:

- Linux Unified Key Setup (LUKS) keys are missing after upgrade prevents node from booting.

**Operations lifecycle**:

- The `iam-ais` subcomponent fails due to a race condition.

## Hotfix 1

Fixed The 1.15.3-gdch.3086-1 hotfix is available. This hotfix fixes the following issues:

**Node OS**:

- Updated the Rocky OS image version to 20260107 to apply the latest security patches and important updates.