This page contains release notes for features and updates to the Compute Engine service.
Latest API version: v1
- VM Deletion Protection is now Generally Available.
europe-west4contains Skylake zones that are now available to all projects and users. See Regions and Zones for more information.
northamerica-northeast1contains Skylake zones that are now available to all projects and users. See Regions and Zones for more information.
- Zonal DNS names are available in Beta. Zonal DNS names are unique to each zone on your internal VPC networks. Global DNS names continue to resolve, but you can enable zonal DNS names to improve the fault tolerance of your applications when they reference instances on your internal VPC network. Read Zonal DNS to learn more.
- NVIDIA® Tesla® P100 GPUs:
- NVIDIA® Tesla® K80 GPUs:
- The OS Login API accepts connections from users who have the Owner or Editor legacy roles. Read Managing SSH keys using the OS Login API to learn more about using OS Login to manage authentication and authorization on Compute Engine instances.
- Added support for Preventing VM Deletion.
- Added support for creating a VM instance from an instance template into Beta.
- Added support for specifying a static internal IP to General Availability. See Reserving a Static Internal IP Address
- Deploying Docker Containers on Compute Engine is now available in Beta.
- Effective today, the Skylake platform no longer incurs a premium charge. Prices for 96 vCPU machine types have been updated to reflect the removal of the Skylake premium. See the price sheet for more information.
- Launched new mega-memory machine types to Beta. See the price sheet for information on how these machine types are billed.
The User Accounts service (previously in Beta) is being discontinued and will stop being supported on February 15, 2018. It is recommended that users transition to the OS Login API in place of the User Accounts service.
Added support for viewing autoscaler logs into Beta. See Viewing Autoscaler Logs for more information.
Windows Server for Containers is now available in Beta as a public image.
- You can associate PTR records with the primary network interface on your instance. Support for PTR records on VM instances is available in Beta. Read Creating a PTR record for VM instances to learn more.
- Added new
asia-south1contains Skylake zones that are now available to all projects and users. See Regions and Zones for more information.
Windows Server, version 1709 is now Generally Available as a public image. This is a Server Core image that is part of the semi-annual release cycle for Windows Server. Use these images to access newer Windows Server features that are not available in the Long-term Servicing Channel releases.
SQL Server 2017 images are now available. These images include Windows Server 2016 with SQL Server 2017 preinstalled.
- Launched new
roles/computeViewerroles to General Availability. For more information on IAM roles, read the Compute Engine IAM documentation.
- You can associate your public SSH keys with your Google account or with individual member accounts in a G Suite organization. When you connect to an instance, those public keys are automatically sent to instances where you have the necessary roles or permissions. This method is an alternative to managing your SSH key pairs in project or instance metadata. This feature is in Beta. Read Managing SSH keys for more information.
- Launched support for Specifying a Minimum CPU Platform to General Availability.
- Added support for sizing recommendations for managed instance groups to Beta. See Applying Sizing Recommendations for Managed Instance Groups for more information.
- You can now copy images from other images, including images that are shared from other Cloud Platform projects. This feature is Generally Available. To learn how to copy images from these sources, see Creating Custom Images.
Added support for disabling external IPs for VM instances using Organization Polices to General Availability.
96-vCPU machine types are now available in Beta on VM instances in specific zones. 96-vCPU machine types can run only on the Skylake platform. Read the list of available regions and zones to see where 96-vCPU machine types are available.
- Added support for selecting specific zones for regional managed instance groups into Beta. See Distributing Instances using Regional Managed Instance Groups
- Added support for Nested Virtualization into Beta.
NVIDIA® Tesla® P100 GPUs are now available in the
europe-west1-dzone in addition to the previously announced zones. Read GPUs on Compute Engine to learn more about the zones where P100 GPUs are available.
NVIDIA® Tesla® K80 GPUs are now available in the
asia-east1-bzone in addition to the previously announced zones. Read GPUs on Compute Engine to learn more about the zones where GPUs are available.
- Billing increments for Google Compute Engine virtual machine instances are reduced from per-minute increments to per-second increments. Additionally, the minimum usage cost for these resources is reduced from 10 minutes to 1 minute. Read the Compute Engine billing model page for details.
NVIDIA® Tesla® P100 GPUs are now available in Beta. Read GPUs on Compute Engine to learn more about the zones where P100 GPUs are available.
NVIDIA® Tesla® K80 GPUs are now Generally Available. Read GPUs on Compute Engine to learn more about the zones where GPUs are available.
- Instance identity verification is now Generally Available for all users and projects. Read Verifying the Identity of Instances to learn how to request and verify signed instance tokens.
- Committed Use Discounts are now Generally Available. To learn more, read the documentation for Committed Use Discounts.
- The Compute Engine Trusted Images policy is now available in Beta. Use this policy to control which boot disk images your project members can access. Read Restricting Access to Images for more information.
Managed Instance Group Updater is now available in Beta. Read Updating Managed Instance Groups for more information.
southamerica-east1contains Broadwell zones that are now available to all projects and users. See Regions and Zones for more information.
- Launched audit logging for Compute Engine to General Availability. For more information, read Viewing Audit Logs.
- Launched new
roles/compute.viewerroles to Beta. For more information on IAM roles, read the Compute Engine IAM documentation.
- The Skylake Platform is now available in zones in the following regions:
asia-southeast1. For a detailed list of zones that support Skylake, see the Regions and Zones documentation.
- You can now create custom images from disks even while they are attached to instances. If necessary, you can create images while the instance is running. Read Creating Custom Images to learn more.
- Added support for specifying a static internal IP to Beta. See Reserving a Static Internal IP Address for more information.
- Introduced new preemptible pricing for local SSD and lowered normal local SSD by up to 63%. For more information, see the Pricing documentation.
- Added support for newer-style health checks that support TCP, SSL, HTTP, and HTTPS health checking to managed instance group autohealing.
- Launched audit logging for Compute Engine to Beta. For more information, read Viewing Audit Logs.
- Added new
europe-west3contains Broadwell zones that are now available to all projects and users. See Regions and Zones for more information.
- You can now use the
gcloudcommand-line tool to connect to instances without external IP addresses. This feature is available in Beta.
- Extended memory is now available in general availibility. For more information, read Adding extended memory to a machine type.
- Instances with service accounts can now request JSON Web Tokens from their metadata servers. Your applications can use these tokens to verify the identity of an instance before transmitting data to the instance. Read Verifying the Identity of Instances to learn how to request and verify signed instance tokens. This feature is available in Beta.
- The Skylake Platform
is now available in the
TCP Proxy Load Balancing is now generally available for all users and projects.
You can now copy images from other images and images that are shared from other Cloud Platform projects. This feature is available in Beta. To learn how to copy images from these sources, see Creating Custom Images.
- Added new
australia-southeast1contains Broadwell zones that are now available to all projects and users. See Regions and Zones for more information.
- Audit logging for the serial console is now available in General Availability.
- The ability to change the service account identity and access scopes for a VM instance is now generally available.
Shared VPC (Previously Cross-Project Networking (XPN)) is now available in General Availability.
NVIDIA® Tesla® K80 GPUs are now available in Beta in the
europe-west1-dzones. Read about GPUs on Compute Engine to learn more about the zones where GPUs are available.
- Added new
europe-west2contains Broadwell zones that are now available to all projects and users. See Regions and Zones for more information.
Launched new Skylake Platform into General Availability.
Launched new feature for Specifying a Minimum CPU Platform to Beta.
Launched new Extended Memory for Custom Machine Types into Beta.
- Added new zone,
us-west1-c. See the Zones and Region documentation for more details.
- Labels are now available in general availability.
- SLES for SAP images are now available as public images. These images are optimized for SAP applications.
- Server Core for Windows Server public images are reduced in size from 50GB to 32GB to reduce the boot disk costs for your instances. By default, these images create boot persistent disks that are 32GB in size. If you require larger boot disks, specify a larger boot disk size when you create your instances or resize the root persistent disk.
- Added new
us-east4contains Broadwell zones that are now available to all projects and users. See Regions and Zones for more information.
64-vCPUs machine types are now Generally Available to all users and projects.
Decoupled labels and tags so that creating either a label or a tag will not create the opposing resource. For example, creating a label will no longer create a tag and vice-versa. For more information, read Relationship between instance labels and network tags.
You can now find information about network tags in the Networking documentation.
- Ubuntu 17.04 Zesty Zapus is available as a public image. You can use this image to create instances.
- Ubuntu 12.04 Precise Pangolin is no longer available as a public image. Optionally, you can upgrade instances that run Ubuntu 12.04 to a newer Ubuntu LTS version.
- Added an API for listing references between resources into Beta.
- Added support for disabling external IPs for VM instances using Organization Polices to Beta.
- Created project quota limits that apply to All Regions.
asia-southeast1contains Broadwell zones that are now available to all projects and users. See Regions and Zones for more information.
Egress firewall rules are now available in Beta to all users and projects.
- TCP proxy load balancing is now available in Beta to all users and projects.
- Added Debian 9 'Stretch' testing image to
debian-cloud-testingproject. For more information, see the Debian Testing documentation.
- Added support for disabling interactive serial console access using Organization Polices into Beta.
- NVIDIA® Tesla® K80 GPUs are now available in Beta in the
us-west1-bzone. Read about GPUs on Compute Engine to learn more about the zones where GPUs are available.
Launched new Committed Use Discounts feature into Beta, which allows you to purchase committed use contracts in return for deeply discounted prices for VM usage. To learn more, read the documentation for Signing Up for Committed Use Discounts.
SSD persistent disks now have increased throughput and IOPS performance, which are particularly beneficial for database and analytics workloads. Instances with 32 vCPUs provide up to 40k read IOPS and 30k write IOPS as well as 800 MB/s of read throughput and 400 MB/s of write throughput. Instances with 16 - 31 vCPUs provide up to 25k read or write IOPS, 480 MB/s of read throughput, and 240 MB/S of write throughput. For complete details about persistent disk performance limits, read Persistent Disk Performance.
Simultaneous reads and writes on SSD persistent disks no longer share the same throughput limits. SSD persistent disks can simultaneously read and write at their individual advertised throughput limits. For more information about simultaneous read and write capabilities, read Persistent Disk Performance for details.
- Cross-Project Networking (XPN) is now available in Beta to all users and projects.
SQL Server Enterprise Edition images are now generally available. You can use these Public Images to run SQL Server Enterprise Edition on various versions of Windows Server. Additionally, you can use SQL Server Enterprise on Compute Engine to create SQL Server Availability Groups.
Volume Shadow Copy Service (VSS) snapshots for persistent disks are now generally available. Use VSS snapshots to create snapshots of persistent disks on Windows instances without detaching the disk or stopping the instance.
- 64-core machine types are now available in Beta to all users and projects.
- Launched new
roles/compute.instanceAdmin.v1role to General Availability. For more information on IAM roles read the Compute Engine IAM documentation.
- NVIDIA® Tesla® K80 GPUs are now available on Compute Engine in Beta. Attach one or more of these GPUs to your instances to accelerate specific workloads and offload tasks from your vCPUs. Add GPUs to your instances or read about GPUs on Compute Engine to learn more.
- Launched the Compute Engine Image Sharing role,
compute.imageUser, to General Availability. For more information, read Sharing Images Across Projects.
- Launched new VM Migration Service to help users migrate VMs into Compute Engine.
SQL Server Enterprise Edition images are now available in Beta. You can use these Public Images to run SQL Server Enterprise Edition on various versions of Windows Server. Additionally, you can use SQL Server Enterprise on Compute Engine to create SQL Server Availability Groups.
Server Core for Windows Server 2016 and Server Core for Windows Server 2012 R2 are now available as public images. Use these images to run Windows Server on smaller instances, to save boot disk space, or to run applications that do not require the complete Desktop Experience.
- The size limit for importing boot disk images increased from 100GB to 2048GB (2TB). See image import requirements for details.
- Added support for autoscaling charts into the Google Cloud Platform Console. Read Viewing autoscaling charts for CPU utilization for more information.
- Interacting with the Serial Console is now generally available.
Added the ability to change the service account and scopes assigned to an instance to Beta.
It is now possible to view audit logs for the serial console. For more information, see Viewing serial console logs.
- Container-VM Image is now called Container-Optimized OS from Google. Container-VM Image has been renamed to Container-Optimized OS to better reflect its focus on container technology and the value it provides. Learn more at Container-Optimized OS from Google Documentation
- Volume Shadow Copy Service (VSS) snapshots for persistent disks are now available in Beta. Use VSS snapshots to create snapshots of persistent disks on Windows instances without detaching the disk or stopping the instance.
- Launched Regional Managed Instance Groups into General Availability.
asia-northeast1contains Broadwell zones that are now available to all projects and users. See Regions and Zones for more information.
Launched new IAM role,
roles/compute.imageUser, into Beta, which allows users to get, list, and use images from another project. For more information, read Sharing Images Across Projects.
- Customer-Supplied Encryption Keys
are now available in the following additional countries:
- New Zealand
- Windows Server 2016 and Windows SQL Server 2016 on Windows Server 2016 are now available as Public Images. You can use these images to create Windows instances.
- Launched a new autoscaling feature that supports queue-based autoscaling into Alpha. You can request access to the feature on the documentation.
- SQL Server images are now generally available. These images include Windows Server with SQL Server preinstalled. The price for SQL Server images includes the licensing cost for SQL Server. This allows you to pay for SQL Server only when you use it with pay as you go billing on a per-minute basis. See the cost for these SQL Server images on the pricing page.
- Added support for new preemptible CPU quotas.
- Launched new Instance Group Updater into Alpha.
- Deprecated previous Instance Group Updater API. Users should transition to using the new Instance Group Updater API instead.
- Instances with 3 TB of total local SSD space are generally available. You can create instances with eight local SSD devices for a total 3 TB of local SSD storage space. See Local SSD limits for available zones and machine types.
- Added new
guestOsFeaturesproperty into Beta, which lets you enable certain guest OS features for your images. For more information, see the documentation for guestOsFeatures.
- SQL Server images are now available in Beta. These images include Windows Server with SQL Server preinstalled. The price for SQL Server images includes the licensing cost for SQL Server. This allows you to pay for SQL Server only when you use it with pay as you go billing on a per-minute basis. See the cost for these SQL Server images on the pricing page.
- SSD persistent disks now have improved IOPS performance. Instances with 16 or more cores can achieve 20,000 IOPS, and instances with 32 vCPUs can achieve 25,000 IOPS with SSD persistent disks of a sufficient size. See the persistent disk performance page for details.
- Lowered prices of preemptible virtual machine instances for predefined and custom machine types. To see the new pricing, see the pricing table for machine types.
- Virtual machine sizing recommendations are now available in Beta.
us-west1region is now available, and offers Broadwell zones
us-west1-b. See Regions and Zones for more information.
Regional managed instance groups are now available in Beta. You can use regional managed instance groups to distribute instances across multiple zones in a region and improve your application availability.
- Shutdown Scripts are now generally available to use with Compute Engine instances. Shutdown scripts allow users to execute commands, on a best-effort basis, right before an instance is terminated or restarted. For more information, see Running Shutdown Scripts.
- Protect data on Compute Engine with your own encryption keys. Customer-Supplied Encryption Keys are now generally available for select countries. You can now also stop an instance with a persistent disk that is encrypted with your own key. Compute Engine is able to restart your instance if you provide the key.
- Released a new feature into Beta where you can enable interactive access to the serial console so you can more easily troubleshoot instances that are not booting properly or that are otherwise inaccessible. See Interacting with the Serial Console for more information.
- Fixed a bug with how Compute Engine accounts for and computes the
metric used to create CPU usage graphs shown in the Google Cloud Platform Console VM
Instances page and Instance Details page. These graphs should now report CPU
usage numbers much closer to the CPU usage inside the guest, and should match
numbers reported by guest tools like
The metadata server has been updated so that it only gives a 60 second advance notice before a maintenance event if you have queried for the
maintenance-eventattribute of that instance at least once since the last migration. Otherwise, Compute Engine assumes that you do not need advance notice of maintenance events and will not provide a notice 60 seconds before an imminent maintenance event. To learn more about the
maintenance-eventattribute, read the Getting transparent maintenance notices section.
You can now specify an instance's internal IP address at instance creation.
The Debian 7 images and Debian 7 backports images are deprecated. Use the latest Debian 8 public image to create new instances or upgrade your Debian 7 instances to Debian 8. If you still require Debian 7, see the advisory message.
CentOS, Debian, and RHEL images version v20160606 or newer now include a new guest environment with the following significant changes:
- Linux guest software is installed from, and hosted in, a Google Cloud
repository, and is updateable with standard package management tools
- Images no longer contain the deprecated
gcimagebundletool. If you need to create image file from a Compute Engine instance, use the instructions for Exporting an Image to Google Cloud Storage.
- Debian images include
unattended-upgradesfor security updates. Security updates are installed daily by default.
- Debian images install Google Cloud SDK from a deb package hosted in a
Google Cloud repository. You can update the
gcloudtool in Debian using the
apt-get update; apt-get dist-upgradecommand instead of the
gcloud components updatecommand.
- Debian images automatically expand root disk partitions up to 2 TB regardless of the boot disk size. Previously, if your disk was over 2 TB, Debian images would not expand the root disk at all.
- Debian images include a compiled
python-crcmodlibrary so that composite objects in Google Cloud Storage work correctly with
- Linux guest software is installed from, and hosted in, a Google Cloud repository, and is updateable with standard package management tools such as
The new Linux guest environment is published on GitHub.
The following Compute Engine IAM roles are now generally available:
For more information, read the IAM documentation.
- Images can now be organized into image families for easier management and use. Image families point to the latest version of an operating system image that is not marked as deprecated. Create an instance by specifying an image family with one of the available public images. To organize your own images into image families, create a custom image.
- The Ubuntu 16.04 LTS Xenial Xerus image is available.
- Health checking is available in Beta for managed instance groups. Use health checking to automatically identify and recreate failing instances in a managed instance group.
- You can now attach more than 16 unique persistent disks to an instance with a predefined machine type. Total persistent disk space per instance is still limited to 64 TB. The number of disks depends on the number of vCPUs that the instance has. This feature is available in Beta. See disk number limits for details.
- Added support for labeling disks, snapshots, and images into Beta.
- Resizeable persistent disks are now generally available to all users and projects. You can increase the size of existing persistent disks even while they are attached to running VM instances.
- Persistent disks larger than 10 TB are generally available. You can now create or resize persistent disks to be up to 64 TB in size. Persistent disks larger than 10 TB can be attached only to specific machine types.
- Launched virtual machine sizing recommendations into Alpha.
- Launched regional managed instance groups into Alpha.
Launched Compute Engine IAM roles into Beta. Learn more about Compute Engine IAM roles.
Added support for custom service accounts into Beta.
- Updated the way that SSH keys work in project metadata and instance metadata. If you manage SSH keys manually, use the new metadata values.
- Added a requirement that account owners who are enabled for user accounts must also be granted permission to act as a service account before they can connect to instances that are enabled for service accounts. Read the documentation for more information.
- Custom machine types are now generally available to all projects and users. Learn more about custom machine types.
- Launched support for changing the machine type of a stopped instance into General Availbility.
- OpenSUSE Leap 42 images are now available to all users and projects starting with image opensuse-leap-42-1-v20160202.
- You can now attach up to eight local SSD devices to each of your instances for a total of 3 TB of local SSD space per instance. Attaching more than 1.5 TB of local SSD space to a single instance is a Beta feature and is available only in some zones. See Local SSD limits for details.
- Persistent disks larger than 10 TB are in Beta. You can now create disks up to 64 TB in size. Persistent disks larger than 10 TB can be attached only to specific machine types.
- You can resize persistent disks to provide more disk space and throughput to your instances. This feature is now available in Beta.
- Launched support for custom machine types into Beta.
- Updated the resource quotas page to reflect that quotas are now listed on the Quotas page in the GCP Console.
- Launched support for changing the machine type of a stopped instance into Beta.
- Launched Instance Labels into Beta.
- Added support for filtering on nested fields in the Compute Engine Beta API.
- Updated activity logs so that the log data
structPayloadin JSON format instead of a
textPayloadin protobuf format. See examples in the Activity Logs documentation for more information.
- Added new us-east1 region. us-east1 contains Haswell zones that are now available to all projects and users. See Regions and Zones for more information.
- The User Accounts service is now
available in Beta. Updates include:
- Release of new
beta-accounts..images that have user accounts enabled.
- Update to quota limits.
- Release of new
- Preemptible instances are now *Generally Availableto all users and projects.
- 32-vCPUs machine types are now Generally Available to all users and projects.
- Instance Groups and Autoscaler are now Generally Available to all users and projects.
- Debian 8 images are now available to all users and projects starting with debian-8-jessie-v20150710.
- Protect data on Compute Engine with your own encryption keys. The Customer-Supplied Encryption Keys feature is now available in Beta for select countries.
Updated the User Accounts API to use a new API endpoint:
- Windows images are now generally available to all users and projects. Commands for managing Windows instances are no longer in beta.
- Released new Windows images, 20150629, that supports service account scopes. This removes the restriction that users must make their startup script publicly-accessible for Windows instances.
- Added new Python and Java script that can programmatically reset a Windows password. See Progammatically generate a username and password for more information.
- Updated the User Accounts service to support the latest Ubuntu images.
- Updated Windows authentication process. Windows images v20150511 and later
will use the new scheme by default.
gcloudwill now generate a random password for Windows login; it is no longer possible to manually set a Windows password through
gcloudbut you can set a custom password in the instance.
- Added preemptible instances that you can create and run at a much lower price than normal instances. For more information about how to use these instances in your Compute Engine project, see the preemptible instances documentation.
- Lowered the price of all machine types in all locations. For more information, see the price sheet.
- Removed support for running
sysprep-oobe-script-*startup scripts on Windows virtual machines. We recommend using
windows-startup-script-*keys as replacements. For more information, see Startup scripts.
- Added documentation for configuring network time protocol (NTP) on virtual machine instances. Make sure you adjust your NTP settings before the upcoming leap second on June 30th, 2015**.
- Updated activity logs so that the format of the log data is provided in protobuf rather than JSON. See examples in the Activity Logs documentation for more information.
- Added new User Accounts feature, available in Alpha. User accounts allow you to create Linux user accounts for your virtual machines.
- Released instance
start()features into General Availability. Additionally, stopped instances no longer count towards your CPU resource quotas. See Stopping an instance and Restarting a stopped instance for more information.
- Upgraded us-central1-b to use Haswell processors. All new virtual machines started in us-central1-b will use Haswell processors by default. Existing instances in us-central1-b have been upgraded from Sandy Bridge to Haswell processors. See Zones and Machine Types for a full list of available zones and processors.
- Added location information about Compute Engine regions. To see specific geographic location of regions, see the Regions & Zones documentation.
- Added support for autoscaling with multiple policies.
- Removed europe-west1-a zone, which was deprecated on October 15, 2014.
- Added shutdown script support on Windows for Windows images *v20150310or later.
- Windows 2012 R2 is now available in Beta to all users and projects.
- Released a security bulletin about CVE-2015-1427. See Security Bulletins for more information.
- Released the RHEL 7.1 image,
rhel-7-v20150311. For a full list of new features , see the RHEL 7.1 Release Notes. To use the new image on Compute Engine, see Using RHEL 7 images.
- For RHEL 7.1 images, Red Hat provides the Kubernetes package to help you manage your containers.
- Added new us-central1-c zone. us-central1-c zone is a Haswell zone that is now available to all projects and users. For more information, see Zones.
- Released new Activity Logs feature in Beta as part of the Cloud Logging service.
- Added new 32-core machine types in Beta. For more information and pricing, see Machine Types and Pricing.
- Released new
moveInstance()feature that moves an instance and its attached disks to another zone. See Moving an instance between zones for more information.
- VPN is now available in Beta. For more information, see the VPN documentation.
- Removed the limit on number of API requests per day for all projects. See API rate limits for more information.
- Added new europe-west1-d zone. europe-west1-d zone is a Haswell zone that is now available to all projects and users. Currently, europe-west1-d zone offers 2.3 GHz Intel Xeon E5 v3 Haswell processors. For more information, see Zones and Machine Types.
- Released new Debian 7 images v20150127 which performs automatic resizing of root persistent disks up to 2 TB. For more information, see Repartition a root persistent disk.
- Added UUIDs to virtual machine instances that can be queried through the dmidecode tool. For more information, see Identifying an instance through the UUID.
- Local SSDs are now in General Availability and can be used by all projects and users.
- Released new instance
start()features in Beta. See Stopping an instance and Restarting a stopped instance for more information.
Note: Currently, virtual machines with local SSDs cannot be restarted but we expect to add this functionality in the future. Additionally, during Beta, stopped virtual machines count toward your available CPU quota but we will remove this restriction before this feature is generally available.
- Added instructions for Importing an AMI image into Compute Engine.
- Ubuntu images are now in General Availability.
Added new Windows images alias to
gcloud compute. You can now specify the latest version of the Windows image by providing the following flag with your instance or disk creation request:
For more information, see Starting a new Windows virtual machine instance.
- Windows 2008 R2 is now available in Beta to all users and projects. Additionally, we've also added support for Microsoft license mobility.
- Added support for local SSD in all zones, *exceptus-central1-b and europe-west1-a (which is deprecated).
gcloudversion 0.9.37 and higher now has support for local SSD flags without an additional repository download. Additionally, the command-line flags have changed for creating local SSDs. For more information, see Local SSD.
- Added support for shutdown scripts in Ubuntu images.
- Released new shutdown script feature in Beta for image versions v20141007 and newer. Shutdown scripts allow users to execute commands, on a best-effort basis, right before an instance is terminated or restarted. For more information, see Shutdown scripts.
- Lowered pricing for persistent SSD and persistent disk snapshots. See Persistent disk pricing for more information.
- Lowered network pricing. See Network pricing for more information.
- Autoscaler is now available in Beta and available to all users and projects. See Autoscaler for more information.
- Added new europe-west1-c zone. europe-west1-c zone is an Ivy Bridge zone that is now available to all projects and users. See Zones for more information.
- Released new local SSDs in Beta phase. This is now available to all users and projects. See Local SSD for more information.
Deprecated europe-west1-a zone.
europe-west1-ahas been deprecated and will be permanently turned down on March 29th, 2015. You should move all resources to
europe-west1-band ensure that you are no longer using any resources in
europe-west1-aafter March 29th, 2015.
We expect that an additional zone,
europe-west1-cwill be available in two weeks, on October 31st, 2014.
For instructions on how to move your instances, see Moving an instance between zones.
- Released new RHEL 7 image,
rhel-7-v20141001. For more information, see Operating Systems.
- For RHEL 7 images, updated the firewall configuration so that all traffic is allowed by default, similar to existing CentOS images.
October 01, 2014 + Lowered pricing for all machine types in all locations. For more information, see the price sheet.
- Released new images,
v20140926, that mitigates additional vulnerabilities in the bash security bug. See Security bulletins for detailed information.
- Added new
utilizationTargetTypeproperty in the API and new
--custom-metric-utilization-target-type CUSTOM_METRIC_UTILIZATION_TARGET_TYPEflag in
gcloud computethat specifies how the target value should be measured, either as a
GAUGEvalue or a
DELTA_PER_MINUTEvalue. This property is required if you are specifying a Stackdriver Monitoring metric. For more information, see the Autoscaler documentation.
gcloud computebehavior so creating a new Windows instance requires an image name and the image project. See Starting a new Windows virtual machine for more information.
gcloud computeinto General Availability. Major changes from the last Open Preview release are:
- Added new
compute/regionproperties that can be used to set a default zone and region. To set the properties, run
gcloud config set compute/zone ZONEand
gcloud config set compute/region REGION.
- Added support for overriding disk auto-deletion during instance deletion.
- Updated the output of commands that mutate resources to be more human-friendly. The --format flag can be used for more verbose output.
- Replaced all
gcloud compute instances gethas been replaced with
gcloud compute instances describe).
- Renamed the
- Added support to the
addressescollection for interacting with global addresses.
- Added new
- The scheduled maintenance for europe-west1-a zone has been cancelled and all zones have now transitioned to using transparent maintenance. This means that virtual machines set to live migrate will no longer be taken offline for maintenance in any zone. For information on transparent maintenance and how to set your virtual machines to live migrate, see the Setting instance scheduling options documentation.
- Added new zone,
asia-east1-c, with transparent maintenance support.
asia-east1-cis now available to all projects and users. See the Zones and Region documentation for more details.
Released new Windows images
windows-server-2008-r2-dc-v20140716with the following updates:
- Allows load balancing for Windows virtual machines that are not in a
zone marked by a
-windowssuffix. See the Load Balancing documentation for more information.
- Fixes a bug where Windows snapshots could not start new instances.
- Allows load balancing for Windows virtual machines that are not in a zone marked by a
Enabled support for Windows virtual machine instances in all zones. Windows instances are no longer limited to Windows-specific zones. For information on starting and using Windows instances, see the Operating Systems documentation. Existing
-windowszones will be inaccessible starting August 15th, 2014 and it is recommended that you restart your instance using the newest Windows image in a non-Windows zone before August 15, 2014.
- Added new zone,
us-central1-f, with transparent maintenance support.
us-central1-fis now available to all projects and users. See the Zones and Region documentation for more details.
- SSD persistent disks are now available in General Availability and open to all users and projects. For detailed information about SSD persistent disks, see Types of persistent disks. For pricing information, see the price sheet.
- Released new images
sles-11-sp3-v20140609to address the OpenSSL security bulletin (CVE-2014-0224) for SUSE Linux Enterprise Server.
- Released new images
v20140605to address the OpenSSL security bulletin (CVE-2014-0224). New images include:
- Released new SSD persistent disks in Limited Preview. SSD persistent disks are also charged at a different rate than standard persistent disks.
- Added new Usage Export feature that lets you export daily and monthly rollup reports about your project's detailed Compute Engine usage.
- 16 core machine types are now in General Availability. For pricing, review the price sheet.
- Added new field to Image resources, named
diskSizeGb, which shows the size of the image when it is restored to a persistent disk, in GB.
Updated default firewall rule names. Default firewall rules are automatically created with every project. These rules were previously named
default-ssh. New projects will have the same default firewalls but with the following new names:
default-allow-internal- Allows network connections of any protocol and port between any two instances.
default-allow-ssh- Allows TCP connections from any source to any instance on the network, over port 22.
Introduced new default firewall rule that will be created with each new project.
default-allow-icmp- Allows ICMP traffic from any source to any instance on the network.
- Updated default Compute Engine API rate limit from 50,000 requests/day to 250,000 requests/day. See API rate limits for more information.
- Introduced new
Metadata-Flavor: Googleheader to replace the
X-Google-Metadata-Request: Trueheader. This also allows users to easily detect if they are running in Compute Engine by querying for the new header. For more information, see Metadata Server.
- Introduced an Asia Pacific region (
asia-east1) and two new supported zones,
- Released new images
v20140408to address the OpenSSL security bulletin (CVE-2014-0160). New images include:
RHEL images have moved to General Availability status and are open to all users and projects.
Note that there is an additional fee for using premium operating systems, including RHEL. Please review the price sheet for more information.
Added new Red Hat Cloud Access feature, which allows users to use their RHEL licenses on Compute Engine virtual machine instances.
Removed support for v1beta16. Please transition to using [v1](/compute/docs/reference/latest/ if you haven't already.
- gcutil Release 1.15.0
- Added feature where gcutil prompts the user to set an initial
Windows password in the
addinstancecommand if the source image is from a Google Windows project.
- Added feature where gcutil prompts the user to set an initial Windows password in the
Introduced sustained use discounts. Sustained use discounts lower the effective price of your instances as your usage goes up. When you use a virtual machine for an entire month, this amounts to an additional 30% discount. For more information, see the price sheet.
Sustained use discounts are effective starting April 1st, 2014.
Windows Server images are now available in limited preview.
Although we do not currently charge for use, you can review the price sheet for the intended Windows Server image pricing.
SUSE images are now generally available and is available for all users.
Note that Compute Engine will start charging for SUSE images on April 1st, 2014. See the price sheet for more information.
Introduced new Replica Pool service, which allows you to create a managed pool of virtual machines based on a reusable template. For more information, see the Replica Pool documentation, or the Replica Pool API reference.
RHEL images are now in open preview with a new image version,
RHEL images are available to all users at no extra cost until April 1, 2014. On April 1, 2014, Compute Engine will start charging for use of these images according to the price sheet.
Released new Debian, CentOS, and Debian Backports images,
- For Debian images, network time protocol (NTP) is now configured to use Google services instead of the public NTP pool.
Updated image packages
- Google Daemon now syncs ssh keys immediately instead of on a per-minute intervals.
- Improved systemd integration.
- Fixed Google Daemon data corruption bug.
- Startup scripts are now downloaded with curl instead of wget.
- Removed harmless warnings.
- Released gcutil 1.14.2.
- Fixed issue where performing
gcutil moveinstanceswith instances with disks whose autoDelete status is set to true would lead to loss of user data.
gcutil moveinstancesis now compatible with Compute Engine API v1 only.
- Fixed issue where performing
- Temporarily disabled support for Advanced Vector Extensions (AVX). Compute Engine has disabled support for AVX due to a stability issue that we are actively investigating. We will re-enable AVX support as soon as we find and fix the root cause.
- SUSE images are now in open preview. This means that SUSE images are available to all users at no extra cost until April 1, 2014. On April 1, 2014, Compute Engine will start charging for use of these images according to the price sheet.
Added ability for creating and deleting a root persistent disk when a virtual machine instance is created or deleted. See the Instances documentation for more information.
Added support for restoring persistent disk snapshots to a persistent disk of a user-specified size.
It is now possible to use the
sizeGbparameter when restoring a snapshot. This can be used to create a persistent disk that is larger than the persistent disk snapshot. See Restoring snapshots to a Larger Size for more information.
Added support for setting the auto-delete state of a read-write persistent disk.
Released gcutil 1.14.0.
- Switched to new, single API call for creating a virtual machine instance with a root persistent disk.
- Added new command,
setinstancediskautodelete, that sets the auto-delete option for persistent disks attached to virtual machine instances.
- Added support for specifying a disk size when creating a disk using a snapshot.
- Decreased the time spent waiting for SSH keys to propagate during initial instance creation from 120 seconds to 10 seconds.
- Added support for Advanced Vector Extensions (AVX) in new virtual machine instances.
All virtual machine instances created after February 11, 2014 have this feature enabled. To check if your virtual machine instance has this enabled, run the following command in your virtual machine instance:
$ cat /proc/cpuinfo | grep avx
If you need to update your instance to use AVX, you must delete and recreate the instance.
- Released new Protocol Forwarding feature. Protocol forwarding allows you to forward traffic to a single virtual machine instance, using a target. instance. Protocol forwarding provides support for these additional features:
Added support for new Target Instance resources, which allows for non-NAT'ed traffic to be forwarded to a single virtual machine instance.
See Protocol forwarding for more information.
Google Compute Engine is now generally available! Users can now feel confident using Compute Engine to support mission-critical workloads with 24/7 support and a 99.95% monthly SLA. The move to general availability also comes with a host of new features and changes, detailed below.
Changes in v1 include (but are not limited to):
New support for custom kernels and removed support for Google-provided kernels
Users can now use custom kernels with their images and no longer need to use Google-built kernels. The Kernels collection has been removed from v1 and all new images will include embedded kernel binaries as part of the image.
Removed scratch boot disks from v1.
All scratch boot disks have been deprecated and we recommend transitioning to using persistent disks. In the v1 API, it is not possible to create a scratch boot disk.
Deprecated *-d machine types.
*-dmachine types have been deprecated and no longer supported. Although you can still create instances with these machine types, we do not recommend this and will eventually remove these machine types completely.
We've introduced a new persistent disk model. Persistent disk performance now scales linearly with the size of the disk. Additionally, we are removing I/O charges for persistent disks completely and lowering the price of persistent disk storage. For more information, review the pricing documentation.
Released new metadata server version v1. The following are new changes with the v1 metadata server:
Requests to the metadata server will now require a security header. All requests to the metadata server will require the following header:
Requests containing the header
X-Forwarded-Forwill automatically be rejected.
Released gcutil v1.12.0.
- Added awareness of deprecated machine types to
listmachinetypesand the machine type prompt when creating instances.
--persistent_boot_diskthe default setting for the
addinstancesubcommand since scratch disks were removed from the v1 API. The
--nopersistent_boot_diskflag can only be specified using the v1beta16 API.
- Deprecated all kernel-related subcommands and flags when using the v1 API.
- Updated gcutil to be distributed with the Cloud SDK.
- Raised the default size of persistent disks to 500GB.
- Made v1 the default API version.
- Added awareness of deprecated machine types to
As part of the Google Compute Engine move to using full disk operating system images, we have made the following changes:
- Released new
backports-debian-wheezyimage, which allows users to access new features and bug fixes from the backports kernel.
- Deprecated Kernels collection.
- Remove all support for kernels from the v1 API.
- Additionally, FreeBSD, SELinux, and CoreOS images now known to be functional on Compute Engine instances with the move to full disk operation system images.
- Released new
Introduced new premium operating systems limited preview program. The new premium OS limited preview program lets you use a SUSE or Red Hat Enterprise Linux (RHEL) images built explicitly for Compute Engine instances. Users who are interested in the program can review the documentation and sign up for the program on the OS page.
Released new Debian 7 and CentOS 6 images,
- New images now contain embedded kernels rather than Google-built kernels. For instructions on how to upgrade you persistent disk to use an embedded kernel, review the documentation. Similarly, you can also upgrade your custom image to use an embedded kernel.
- New images allow you to use dmidecode to determine if you are running on Google Compute Engine. See the documentation for more information.
Deprecated the Kernel resource. Google will no longer provide custom kernels and will instead use community-provided kernels in Google-provided images.
Added new instance migration and transparent scheduled maintenance features. Google Compute Engine now offers transparent scheduled maintenance in
us-central1-b; these zones will no longer go offline for scheduled maintenance and Google Compute Engine will automatically move your instances out of the way of any scheduled maintenance activity. For more information, see maintenance events.
Added new gcutil release 1.11.0.
- Added a new subcommand,
gcutil whoami, that prints out the email of the currently-authenticated user to standard out.
- Added two new scope aliases: datastore and userinfo-email.
- Added flags to
gcutil addinstanceand a new subcommand,
gcutil setscheduling, for controlling instance scheduling parameters.
- Disabled host key checking for commands that rely on ssh because there is no secure channel to pass the host key to the client for the first time.
- Added a new subcommand,
Marked all Debian 6 images as deprecated.
Marked Debian 7 images older than
- Deprecated us-central2-a zone.
us-central2-ahas been deprecated and will be permanently turned down by December 31st, 2013. You should move all resources to
us-central1-b(after November 11, 2013) and ensure that you are no longer using any resources in
us-central2-aafter December 31st, 2013.
- Added new kernel,
gce-no-conn-track-v20130813, and images
gce-no-conn-track-v20120813kernel is identical to
gce-v20130813 kernelexcept that connection tracking is no longer enabled.
v20130926will use the new
gce-no-conn-trackkernel. To use a kernel with connection tracking turned on, specify the
--kernelflag with a previous kernel version, such as
Reduce duration of two upcoming maintenance windows for
us-central1-bzones. The new maintenance window durations are as follows:
us-central2-a: Oct 12, 2013 12:00:00 PM - Oct 22, 2013 10:00:00 AM
us-central1-b: Nov 2, 2013 12:00:00 PM - Nov 10, 2013 12:00:00 PM
Released gcutil 1.9.1.
- Fixed a bug in which the tilde in the authentication file path was not being expanded properly.
Added new features to load balancing:
- Removed zone quotas.
- Added new regional quotas.
- Updated the global default quotas with new default limits.
addresses().userfield from a string to a list and renamed the field to
- Added new
setBackupmethod to set backup target pools for existing primary target pools.
- Updated TargetPools resource representation to describe backup pools, failover ratios, and session affinity.
Released gcutil 1.9.0.
- Added new
--failover_ratioflags for the
- Added new
- gcutil now outputs tables thats respect the terminal width. This
feature can be turned off using the
gcutil deleteinstancewith the
--forceflag now requests users to explicitly provide
--[no]delete_boot_pdif any of the instances have a boot disk.
Stopped allowing cross-project resource references, such as the ability to create a disk from a snapshot in another project. Previously, it was allowed for projects whose access control lists (ACLs) allowed it, such as situations where multiple projects were owned by one user.
- Released gcutil 1.8.4.
- Fixed an issue whereby reserved IP addresses were not preserved in the
- Bug fixed where global flags were not being displayed on
- Updated gcutil help text.
- Fixed an issue whereby reserved IP addresses were not preserved in the
- Added new Debian images
- Updated images to use latest kernel.
- Updated images to use latest gcutil too.
Removed support for v1beta14.
(Updated 09/09/2013) Removed support for cross-region external IP address assignment.
- Added support for differential snapshots.
- Added information on how to send email using SendGrid.
- Added new CentOS image
v20130813with the following updates:
- Updated image to use the latest kernel.
Updated image to use the latest gcutil tool.
Added new kernels
v20130813with the following updates:
- Added multiqueue support.
- Fixed an issue in scheduler that impacted Hadoop.
- Added backport pvclock enlightment for softlockup detector.
Launched new load balancing service. Google Compute Engine has launched a load balancing feature that lets you distribute traffic across your instances. Load balancing is especially useful for supporting heavy traffic to your instances and to provide redundancy to avoid failures. For more information, visit the load balancing documentation. Additionally, you can review the load balancing reference documentation.
Released gcutil 1.8.3
Added new prompt to select a persistent or scratch boot disk when using
Changed naming of persistent boot disks that are created during instance creation from
- Added prompt to delete attached persistent disk when using
- Added support for load balancing.
Added source code for custom tools that Google Compute Engine images uses, onto GitHub. The list of tools include:
- Image Bundle - Creates an image file our of a disk attached to a virtual machine instance.
- Google Startup Scripts - Scripts and configuration files that set up a Linux-based image to work smoothly with Google Compute Engine.
- Google Daemon - A service that manages user accounts, maintains SSH login keys, and syncs public endpoint IP addresses.
Added new Debian and CentOS images
v20130723, with the following updates:
- Added latest gsutil version which addresses issues where gsutil was not working properly.
- Fixed typo which causes erroneous startup-script-url error.
- Marked kernels older than
Marked deprecated kernels
gce-v20120912and older as
OBSOLETE. For a list of kernels and their deprecation states, run the following command:
gcutil --project=<project-id> listkernels
- Added bursting for
f1-microinstances. See machine types for more information.
- Added ability to reset an instance through the API. Review the
documentation for resetting instances for more
information, or review the
- Released gcutil 1.8.2.
- Added new
gcutil resetinstancecommand that allows resetting virtual machine instances.
- Fixed region detection when releasing addresses from multiple regions.
- Fixed aggregated resource listing with
- Added new
- Fixed the usage help string for
- Added new Debian images
Added the following updates for Debian 6 and 7 images
- Updated gsutil to 3.31 and gcutil to 1.8.1.
- Disable IPv6 by default via /etc/sysctl.d, for optimal user experience. Google Compute Engine does not currently support IPv6.
Added the following updates for Debian 7 image
- Upgrade pre-installed packages to Debian 7.1, incorporating security updates and miscellaneous important bug fixes. For more information, see the Debian announcements.
- Added new images
- Patched new kernel version
gcg-3.3.8-201305291443to address vulnerability in previous kernels. See Security Bulletins for more information.
- Fixed kernel warning printed on boot about virtio net multiqueue.
- Made ext4 kernel fixes (for
- Increased default per-project total disk quota to 1TB.
- Updated gcutil:
- Updated documentation for
gcutil moveinstancesto provide a warning of possible failures during the moving process.
- Improved error detection in the
- Fixed behavior where gcutil attempted to use existing persistent disk
when recreating an instance with the same name and the
- Machine type prompts in gcutil now provides a description of the
machine types and
gcutil listimageswill now only display the name and description of images.
- Updated documentation for
Google Compute Engine is available for open signups! We're excited to announce that Google Compute Engine is now available for open signups and anyone can sign up for the service. For signup instructions, see the signup page.
Released new API version v1beta15. v1beta14 is now deprecated and customers should switch to v1beta15. v1beta14 will remain available until August 15, 2013 and v1beta13 will be discontinued on May 31, 2013. Changes in v1beta15 include:
Introduced new region scope and regional resources.
Added new *regional resource URIsto access regional resources, in the form:
For example, to access regional reserved IPs, use the following regional URI:
Updated reserved IP addresses to a regional resource.
External static IPs are now referred to as reserved IP addresses and are no longer a global resource. Reserved IPs are now a regional resource that can be managed through the Addresses collection.
You can also provision, promote, and release external IP addresses through the Addresses collection, without having to manually request one. For more information, see the Reserved Addresses documentation.
Converted machine type resources to per-zone resources.
To use a machine type, you must now specify the zone in which that machine type lives:
Changed method of creating Snapshot resources to use a custom verb on the Disk resource.
To create a Snapshot resource, you must now make a request to the following URI:
Snapshots are still accessible by making requests to the Snapshot collection.
Removed ability to assign an internal IP address.
internalIpfield on a virtual machine instance is now read-only and you can no longer manually assign internal IPs to your instances. Google Compute Engine will assign internal IPs automatically.
Added a number of new features.
- Added new Routes collection that lets you set up and manage a virtual machine's routing table.
- Added ability to reserve and release static IPs, and to promote ephemeral IPs to static IPs.
Added the ability to request aggregate lists for per-zone and per-region resources. You can request aggregate lists for the following resources:
- Instance resources
- Disk resources
- Address resources
- Machine type resources
For example, you can list instances across all zones by making a request to the following URI:
Introduced new shared-core machine types.
Shared-core machine types are more cost-effective for running applications that don't require a lot of resources. New available machine types are
Updated maximum total persistent disk size that can be attached to a machine type.
Standard, high-memory, and high-CPU machine types now have an updated maximum total disk size of 10 TB. See machine types for more information.
Updated billing model for instances. Google Compute Engine has updated our billing model so that instances are billed based on per-minute usage. All instances that run for 10 minutes or less will be charged for 10 minutes of usage. After the first 10 minutes, usage is charged on a per-minute basis.
Added new images and kernels
Removed Google-specific repositories from images. The only packaged repositories configured in images are now the Debian archive. Google Compute Engine still installs Google-specific packages at build time but removed Google- specific repositories for various reasons.
Removed default installation of the
Changed log location of startup script output to
/var/log/startupscript.log. Also, added startup script log output to the instance's serial port console so you can also run
gcutil getserialportoutputto retrieve startup script log information.
Improved instance creation and deletion time for Debian.
Fixed issue preventing startup script specified in metadata to be downloaded from Google Cloud Storage.
dist-upgradefrom starting on instance boot.
Released gcutil 1.8.0.
- Added support for v1beta15 Google Compute Engine API. (addresses, regions, per-zone machine types, aggregated lists).
gcutil configcommand, an alias for
- When prompting the user to select an image, gcutil will include standard images (CentOS, Debian).
- With v1beta15 API, gcutil will use aggregated list API call by default. Aggregated list method will aggregate all resources across all scopes in which the resource of that type exist (for example, aggregated list of instances will list instances in all zones).
- Users can specify image from the standard project by specifying image
name prefix. For example:
gcutil addinstance my-instance --image=debian-7.
- When moving instances using
gcutil moveinstances, if some of the instances depend on deprecated resources (image, kernel), gcutil will warn before it proceeds with the migration (migration would fail). New flag
--replace_deprecatedwill create instances in the destination zone with dependencies on deprecated resources updated to recommended replacement resources.
Listcommands will display all resources by default. Number of resources listed may be limited using
--fetch_all_pagesflag is now deprecated.
- Improved display of images and kernels list. By default, only newest
kernels/images will be displayed when listed or when user is prompted
to select an image or kernel. Use
--old_kernelsto list all images or kernels, respectively.
- When listing imges, the standard images (CentOS, Debian) will be
listed in addition to images from the specified project. To list
images in the specified project only, use
- When prompting user to select a machine type, gcutil displays machine type description in addition to the name.
- Removed support for v1beta13 Google Compute Engine API.
gcelib is no longer available and if you haven't already, we strongly encourage users to transition to the Google APIs Python Client Library.
Released new Debian images. Google Compute Engine is happy to announce that Debian images for Google Compute Engine are now available for your instances. To view a list of Debian images available to your project, run the following gcutil command:
gcutil --project=debian-cloud listimages
For information about Debian images, see the Debian wiki.
Similarly, you can see a list of CentOS images like so:
gcutil --project=centos-cloud listimages
Deprecated gcel images. gcel images are now deprecated and we encourage users to transition to either Debian or CentOS images.