Archived release notes

Generally available: The accelerator-optimized A3 machine type is now available on Compute Engine. The A3 machine type has NVIDIA® H100 80GB GPUs attached, which can be used to support your large artificial intelligence (AI) models, machine learning (ML), and high performance computing (HPC) workloads. The A3 machine type is available in the following regions and zones:

• APAC
  • Singapore: asia-southeast1-b,c
• Europe
  • Netherlands: europe-west4-b.c
• North America
  • Iowa: us-central1-a,c
  • Virginia: us-east4-a,c
  • Ohio: us-east5-a

Generally available: You can rename an existing VM using the Google Cloud console, gcloud CLI, and REST. For more information, see Rename a VM.

Generally available: NVIDIA L4 GPUs are now available in the following additional regions and zones:

• Seoul, South Korea (asia-northeast3-a)
• Moncks Corner, South Carolina (us-east1-c)

For more information about using GPUs on Compute Engine, see GPU platforms.

Preview: The following quotas and metrics are now available to help you monitor the usage and limits for Compute Engine concurrent operation quotas:

• Quotas for global concurrent operations (metric - compute.googleapis.com/global_concurrent_operations):
  • Concurrent global operations per project
  • Concurrent global operations per project operation type
• Quotas for regional concurrent operations (metric: compute.googleapis.com/regional_concurrent_operations):
  • Concurrent regional operations per project
  • Concurrent regional operations per project operation type

For more information, see Concurrent operation quotas.

Preview: Managed instance groups (MIGs) let you create pools of suspended and stopped virtual machine (VM) instances. You can manually suspend and stop VMs in a MIG to save on costs, or use suspended and stopped pools to speed up scale out operations of your MIG.

For more information, see Work with suspended and stopped VMs in a MIG.

Generally available: The following location and scale enhancements for Persistent Disk Asynchronous Replication are generally available:

• Larger disk capacity: the maximum disk size has increased from 2 TiB to 5 TiB.
• Faster replication rate: the disk data replication rate has increased from 100 MB/min to 250 MB/min.
• Expanded regional support: PD Async Replication is available in 15 additional regions across Europe, APAC, and North America. For the full list of available regions, see Supported region pairs.

Preview: When creating or modifying an on-demand reservation, you can configure reservations to be automatically deleted at a specific date and time. Automatically deleting reservations makes it easier to prevent charges from unused reservations when you no longer need them.

For more information, see the documentation for creating on-demand reservations.

Preview: In a managed instance group (MIG), you can turn off repairs to inspect failed and unhealthy VMs, to implement your own repair logic, or to monitor the application health without triggering repairs by MIG. For more information, see Turn off repairs in a MIG.

Generally available: A replica recovery checkpoint of a regional Persistent Disk volume represents the most recent crash-consistent point in time of the fully replicated disk. For disks that are not fully replicated, you can use the checkpoint to create disk snapshots from an incomplete zonal replica. You can create and use these snapshots to recover disk data in the rare scenario where your synced replica goes down before your incomplete replica catches up.

Learn more about Regional Persistent Disk replica recovery checkpoints and how to use checkpoints to recover a degraded disk.

The h3-node-88-352 sole-tenant node type is now Generally Available.

Generally available: When assigning a custom queue count for the receive and transmit queues for a vNIC, under certain conditions, you can configure a number of custom queue counts that exceeds the number of vCPUs allocated to the VM.

Preview: Advanced maintenance control for sole-tenancy lets you control planned maintenance events for sole-tenant node groups and minimise maintenance-related disruptions. This feature is available only for sole-tenant node groups. To use this feature with your existing virtual machines, you must first move your VMs to sole-tenant node groups that have advanced maintenance control enabled.

The advanced maintenance control for sole-tenancy feature lets you:

• Check for maintenance events scheduled for a sole-tenant node 28 days in advance.
• Trigger maintenance immediately or schedule it for later. Note that if you trigger maintenance immediately, the maintenance takes place within 24-hours from the time you trigger the request.

For more information, see Advanced maintenance control for sole-tenancy.

Preview: Hyperdisk Balanced is now available in preview with H3 VMs. Hyperdisk Balanced is a good fit for a wide range of use cases such as LOB applications, web applications, and medium-tier databases that don't require the performance of Hyperdisk Extreme. For more information, see About Hyperdisk.

Preview: Project zonal metadata is custom project metadata that you can set exclusively for VMs in a specific zone in a project. Project zonal metadata helps you with fault isolation and provides greater reliability. By setting custom project zonal metadata, you gain more control over the project metadata for your VMs and limit the impact of any incorrect metadata updates to VMs within the specific zone.

Learn more about VM metadata and how to set custom project zonal metadata.

Generally available: c3d-standard, c3d-highmem, c3d-highcpu, and c3d-standard-lssd machine types for general-purpose C3D VMs are generally available.

Generally available: C3 VMs support Compute Engine flexible committed use discounts (CUDs).

Compute Engine flexible CUDs allow you to commit to a minimum hourly spend amount and use vCPUs and/or memory in any of the projects within your Cloud Billing account, across any region, and belonging to any eligible machine types. Learn more about Compute Engine Flexible CUDs and how to purchase flexible commitments.

If you want to modify a future reservation request using the Compute Engine API, the paths query parameter is deprecated. Instead, use the updateMask query parameter.

For more information, see Modify future reservation requests.

Preview: You can now use workforce identity federation with OS Login.

Preview: The following metrics are now available to help you monitor your Persistent Disk and Hyperdisk volume performance:

• Average I/O latency (compute.googleapis.com/instance/disk/average_io_latency)

• Average I/O queue depth (compute.googleapis.com/instance/disk/average_io_queue_depth)

To learn more about these metrics and how to view them, see Review disk metrics.

Generally available: You can configure stateful IP addresses in a managed instance group. Stateful IP addresses are preserved when VM instances in the group are repaired, updated, and re-created. For more information, see Configuring stateful IP addresses in MIGs.

Generally available: H3 VMs, designed for compute-intensive high performance computing (HPC) workloads, are now generally available. For more information, see H3 machine series.

When you install the Ops Agent on a Compute Engine VM by using the Observability tab on a Compute Engine VM details page, the agent is now installed with an Ops Agent OS policy. This installation method replaces the prior set of manual steps. For more information, see Installing the agent by using the Google Cloud console.

Generally available: NVIDIA L4 GPUs are now available in the following additional regions and zones:

• APAC
  • Seoul, South Korea (asia-northeast3-b)
• Europe
  • St. Ghislain, Belgium (europe-west1-b)
  • Frankfurt, Germany (europe-west3-b)
• North America
  • Council Bluffs, Iowa: (us-central1-c)
  • Las Vegas, Nevada (us-west4-a,c)

For more information about using GPUs on Compute Engine, see GPU platforms.

Generally available: NVIDIA L4 GPUs are now available in the following additional regions and zones:

• Singapore(asia-southeast1-a)

For more information about using GPUs on Compute Engine, see GPU platforms.

Creating a reservation or future reservation request by using an instance template that specifies an A2, C3, or G2 machine type causes errors or problems with consumption. For more information, see Known issues.

Preview: c3d-standard, c3d-highmem, c3d-highcpu, and c3d-standard-lssd virtual machines are available in the following regions:

• Council Bluffs, Iowa, North America, us-central1
• Moncks Corner, South Carolina, North America, us-east1
• Ashburn, Virginia, North America , us-east4
• St. Ghislain, Belgium, Europe, europe-west1
• Eemshaven, Netherlands, Europe, europe-west4
• Jurong West, Singapore, Asia, asia-southeast1

See the General purpose machines document for details.

Preview: You can now view the organization-wide patch status dashboard and OS policy compliance reports by using VM Manager.

Preview: Compute Engine API now enforces the Filtered list cost overhead quota, which limits the number of resources to be filtered out from server-side *.list and *.aggregatedList methods.

The quota is charged against the following metrics:

• Global: compute.googleapis.com/filtered_list_cost_overhead
• Regional: compute.googleapis.com/filtered_list_cost_overhead_per_region

For more information, see Rate quotas and best practices for list filtering.

The Google Cloud console labels for OS patch management and OS configuration management on VM Manager pages have been renamed to Patch and OS policies respectively.

Generally available: Autohealing in managed instance groups (MIG) supports regional health checks. To reduce cross-region dependency or to achieve data residency, use a regional health check. For more information, see Set up an application health check and autohealing.

Generally available: Instance templates are available as both regional and global resources. To reduce cross-region dependency or to achieve data residency, use a regional instance template to create virtual machines (VM), managed instance groups (MIG), or reservations. For more information, see Regional and global instance templates.

Generally available: Dammam, Kingdom of Saudi Arabia, Middle East me-central2-a,b,c has launched with E2, N2, N2D, and T2D VMs in all three zones. See the Dammam region access document to learn more.

Preview: Snapshot settings are centralized configuration parameters for all snapshots in a project. You can use snapshot settings to customize the default storage location for all future snapshots in your project. By enabling you to do this, snapshot settings remove the need for you to manually specify a storage location during each individual snapshot creation.

Learn more about snapshot settings and how to set the default storage location for a project using snapshot settings.

Generally available: The Red Hat Knowledgebase provides you with access to articles, solutions, product documentation, and community discussions for Red Hat products.

You can now access the Red Hat Knowledgebase by using single-sign-on (SSO) through the Google Cloud console from your Red Hat Enterprise Linux (RHEL) VMs. For more information, see Access Red Hat Knowledgebase.

Generally available: You can create C3-standard VMs with Local SSD attached using new machine types, for example c3-standard-44-lssd. For more information, see Choosing a valid number of Local SSDs.

You can manage future reservations using the Google Cloud console. Future reservations provide a high level of assurance to obtain important or difficult-to-obtain capacity in advance.

For more information, see the following pages:

• Create future reservation requests for a single project
• Create shared future reservation requests
• Delete future reservation requests
• Modify future reservation requests
• View future reservation requests

Generally available: NVIDIA L4 GPUs are now available in the following additional regions and zones:

• APAC
  • Taiwan (asia-east1-a,c)
  • Tokyo (asia-northeast1-a,c)
  • Singapore(asia-southeast1-c)
• Europe
  • Belgium (europe-west1-c)
  • London (europe-west2-a,b)
• North America
  • Northern virginia (us-east4-c)
  • Dalles (us-west1-c)

For more information about using GPUs on Compute Engine, see GPU platforms.

Preview: You can use future reservations to reserve resources at a specific date and time in the near future.

Future reservations are useful for obtaining capacity for future peak-demand events or highly-demanded resources. For more information, see About future reservation requests.

Preview: You can reduce network latency between VMs by using compact placement policies to specify the maximum distance between VMs. Use compact placement policies to optimize workloads with frequent communication across VMs—for example, high-performance computing (HPC), machine learning (ML), or database server workloads. You can keep VMs in the same rack, across adjacent racks within the same cluster, or across adjacent clusters.

For more information, see Reduce latency by using compact placement policies.

Generally available: Berlin, Germany, Europe europe-west10-a,b,c has launched with E2, N2, N2D, and T2D VMs available in all three zones.

Generally available: When a managed instance group (MIG) repairs a failed or an unhealthy VM, you can apply the latest instance template and per-instance configuration to recreate the VM instead of applying the configuration originally used to create the VM. For more information, see Apply configuration updates during repairs.

Generally available: Hyperdisk Throughput is now available in the following additional regions and zones:

• Council Bluffs, Iowa: us-central1
• Moncks Corner, South Carolina: us-east1
• Ashburn, Virginia: us-east4-b, c
• Eemshaven, Netherlands: europe-west4-a, c
• Jurong West, Singapore: asia-southeast1
• Mumbai, India: asia-south1-a

Generally available: The Ops Agent (version 2.38.0 and later) now supports the automatic tracking of GPU usage metrics reported from the NVIDIA Management Library (NVML) for Linux virtual machine instances that have attached NVIDIA GPUs.

Through an available integration with NVIDIA's Data Center GPU Manager (DCGM), you can also track metrics such as Streaming Multiprocessor (SM) block utilization, SM occupancy, SM pipe utilization, PCIe traffic rate, and NVLink traffic rate.

For more information, see Monitoring GPU performance on Linux VMs.

Generally available: You can use Cloud Monitoring to monitor the consumption of your reservations and set custom alerts. For more information, see Monitor consumption of reservations.

Generally available: If a host error occurs on a VM, you can control how much time Compute Engine spends recovering Local SSD data with the Local SSD recovery timeout setting. For more information, see Local SSD data persistence.

Generally available: Use the new distribution shape ANY SINGLE ZONE in a regional Managed Instance Group (MIG) to automatically select a single zone that has available resources within your quota. Recommended for workloads that require low latency, high-bandwidth connections between VMs or when you want to avoid inter-zone network traffic costs.

Preview: You can create C3 VMs with Local SSD attached using new machine types (-lssd). For more information, see Add a local SSD to your VM.