This is a technical changelog for the Hetzner Cloud platform. To get updated about changes automatically, you can subscribe to our RSS feed
To protect data from getting deleted by accident, Cloud Console now matches the behavior of the S3 API and prohibits the deletion of Buckets, unless:
To empty a large Bucket, we recommend deleting the data via lifecycle policies rather than using an S3 compatible tool. You can find two example lifecycle policies here.
The ISO openSUSE MicroOS Snapshot 2025-04-22 (IDs 119712 (x86) & 119713 (arm)) is now available as ISO for all Cloud Servers.
The newest Fedora release with version 42 is now available as
fedora-42 (IDs 232895138 (x86) & 232895264 (arm)) to all Cloud Servers as a Rapid Deploy Image.
The ISO Talos Linux 1.9.5 (IDs 119702 (x86) & 119703 (arm)) is now available as ISO for all Cloud Servers.
Talos Linux ISOs with customizations applied are called "models". To differentiate them, each model has a unique schematic ID.
We provide the model with the schematic ID ce4c980550dd2ab1b17bbf2b08801c7eb59418eafe8f279833297925d67c7515 (Hetzner metadata support + qemu-guest-agent).
For more information about the schematic ID, please see the Talos image factory documentation.
We have previously announced the actions list endpoint deprecation in July 2023. We have now changed our decision in parts.
The endpoint GET /v1/actions implements two different functions:
GET /v1/actions)GET /v1/actions?id=)The previous announcement mentioned that the whole endpoint would be removed.
We have now decided to keep the GET /v1/actions?id= functionality available.
We have removed the possibility to list all actions (GET /v1/actions).
All query parameters related to this have been removed from the API.
The API now returns a 410 Gone status code and links to this announcement.
If you need to list actions without knowing their IDs, please use the "Resource Action endpoints" instead.
From now on, after deleting a Bucket, the same Bucket name can be reused immediately when it is created in a project of the same owner. Please see our documentation for more details.
The API no longer returns for the following fields:
included_traffic of a server_typetraffic in the GET /v1/pricing endpoint responseA new App is available for your servers:
coolify, IDs 197238872 (x86) & 197238880 (arm))Note: To give customers more time to make the necessary changes on their end, we have extended the deadline for this deprecation.
Starting on 30 October 2024, it will no longer be possible to use custom MAC addresses for IPv4 connectivity on the public network interface. Instead, servers will have to use the MAC address that was assigned to their public network interface to send and receive IPv4 packets. Packets sent with a different source MAC address will be dropped.
For public IPv6 connectivity as well as private network interfaces, this was already enforced.
If you have changed the MAC address on the public network interface of your cloud server, please change it back to the MAC address assigned to the interface. If your server is configured so that the public network interface is part of a bridge with other interfaces, please consider switching to a routed setup.
Customers affected by this change were notified via email.
The old server plans with shared Intel vCPUs (CX11, CX21, CX31, CX41, CX51) are no longer available for order. Existing servers are not affected by this deprecation and they will continue to work. If you want to migrate an existing server to one of the new server plans with Intel vCPUs, you can use the rescale option.
These server types are not available through the API List endpoint. This also affects any usage "by name".
Applies to server types:
1)3)5)7)9)The Hetzner provider in current versions of cluster-autoscaler has a bug and relies on the CX11 server type, which we will remove from our ordering options on 6 September 2024.
If you try to use the cluster-autoscaler provider after that date, you will see the following error messages:
mixed_nodeinfos_processor.go:160] Unable to build proper template node for draining-node-pool: failed to create resource list for node group draining-node-pool error: failed to get machine type cx11 info error: server type not found
static_autoscaler.go:387] Failed to get node infos for groups: failed to create resource list for node group draining-node-pool error: failed to get machine type cx11 info error: server type not found
The following versions of cluster-autoscaler are affected:
We depend on the Kubernetes community and the maintainers of cluster-autoscaler to release new versions. We expect that new official versions are released at the end of September.
To bridge the gap until the Kubernetes community releases the new versions, we published alternative container images of cluster-autoscaler that include a patch for the bug. You can use these in your deployment, but we will remove them one month after new official cluster-autoscaler versions become available. We will not provide any other patch releases on this container image repository. Please switch back to the official images as soon as possible.
docker.io/hetznercloud/cluster-autoscaler:v1.28.6-hcloud1 (Build Commit)docker.io/hetznercloud/cluster-autoscaler:v1.29.4-hcloud1 (Build Commit)docker.io/hetznercloud/cluster-autoscaler:v1.30.2-hcloud1 (Build Commit)docker.io/hetznercloud/cluster-autoscaler:v1.31.0-hcloud1 (Build Commit)To prevent disruptions for existing users of the provider, we will keep the CX11 server type available for these accounts. We will remove that prolonged access to the CX11 server type two weeks after the Kubernetes community releases new versions of cluster-autoscaler.
Starting on 30 September 2024, it will no longer be possible to use custom MAC addresses for IPv4 connectivity on the public interface. Instead, servers will have to use the MAC address that was assigned to their public network interface to send and receive IPv4 packets. Packets sent with a different source MAC address will be dropped.
For public IPv6 connectivity as well as private network interfaces, this was already enforced.
If you have changed the MAC address on the public network interface of your cloud server, please change it back to the MAC address assigned to the interface. If your server is configured so that the public network interface is part of a bridge with other interfaces, please consider switching to a routed setup.
We will also contact customers affected by this change via email.
Update 27 September 2024: We have extended the deadline. You can find the updated entry here.
The new location Singapore (sin) is now available for all products. The following server types are available:
cpx11, cpx21, cpx31, cpx41, cpx51ccx13, ccx23, ccx33, ccx43, ccx53, ccx63The datacenter sin-dc1 was also added to the API and belongs to the sin location.
The network zone ap-southeast was also added to the API. It currently includes the sin location.
The API now returns null for the following fields instead of their previous values:
included_traffic of a server_typetraffic in the GET /v1/pricing endpoint responseThese fields will be fully removed from the API on 2024-11-05.
Please refer to the API documentation for more details.
The following app(s) now use Ubuntu 24.04 as base-image (previously 22.04):
gitlab (IDs 40093435 (x86) & 105888988 (arm))jitsi (IDs 40093140 (x86) & 105887536 (arm))The Cloud API has been updated to provide a better insight and more flexibility for displaying the pricing of traffic for servers and load balancers.
These API changes will be performed in three steps:
The pricing arrays in the responses of our GET /v1/server_types and GET /v1/load_balancer_types endpoints have been extended:
included_traffic representing how much of free outgoing traffic is included for a tariff in a certain location in bytesprice_per_tb_traffic, representing the cost per TB of traffic after the included traffic has been used upThe matching values in responses to the following endpoints are updated accordingly:
GET /v1/pricingGET /v1/serversGET /v1/servers/{id}GET /v1/load_balancersGET /v1/load_balancers/{id}In addition to these new properties, we will deprecate existing properties currently representing these values. These will be set to return null on 2024-08-05:
included_traffic of a server_typetraffic in the GET /v1/pricing endpoint responseThese fields will be fully removed from the API three months later on 2024-11-05.
Please refer to the API documentation for more details. Review and update any scripts or tools that make use of these API endpoints and fields.
We have updated our integrations for these new fields. The changes are available in the following versions:
The image fedora-38 (IDs 107768015 (x86) & 107768016 (arm)) is no longer available when creating new servers.
In order to save bandwidth, we now allow for gzip and brotli
compression to be requested by sending an Accept-Encoding header with
your API request.
We don't expect any complications, but this change could lead to unexpected behaviour if your client requests compression whilst not handling the decompression of the response as expected. Our official integrations will continue to work with this, but we can not ensure full compatibility for third-party clients and integrations.
We will roll out this change in increasingly longer time periods so you can observe if all your implementations can handle the new behaviour or if you have to make changes:
The change will be permanently effective from 26th August 2024 08:00 UTC.
We developed a new system for managing private networks in the recent months and are now in the rollout phase. Starting today, private networks in newly created projects will use the new system. With this new system, private network actions related to progress and status information will behave more accurately.
If you encounter any issues, please let us know by submitting a support ticket.
In the upcoming days, existing private networks will be migrated to the new system. We do not expect any downtime or disturbances for customers.
Four new server types with shared Intel vCPUs are now available:
104)105)106)107)Learn more about these instances in this news article
Starting on 06 September 2024, the old server plans with shared Intel vCPUs will no longer be available for order. Existing servers are not affected by this deprecation and they will continue to work. If you want to migrate an existing server to one of the new server plans with Intel vCPUs, you can use the rescale option.
These server types will not be available through the API List endpoint after the announced date. This also affects any usage "by name".
Applies to server types:
1)3)5)7)9)The following app(s) now use Ubuntu 24.04 as base-image (previously 22.04):
photoprism (IDs 101588775 (x86) & 105888330 (arm))prometheus-grafana (IDs 73721460 (x86) & 105888142 (arm))owncast (IDs 101588646 (x86) & 105888084 (arm))rustdesk (IDs 102456737 (x86) & 105888087 (arm))wireguard (IDs 80379320 (x86) & 105888135 (arm))The following app(s) now use Ubuntu 24.04 as base-image (previously 22.04):
docker-ce (IDs 40093247 (x86) & 105888141 (arm))lamp (IDs 40093059 (x86) & 105888091 (arm))nextcloud (IDs 40093190 (x86) & 105888089 (arm))wordpress (IDs 40093134 (x86) & 105888096 (arm))The following app(s) now use Ubuntu 22.04 as base-image (previously 20.04):
jitsi (IDs 40093140 (x86) & 105887536 (arm))The response field deprecated of all endpoints that return "ISOs" is
removed as announced.
Please use the new field deprecation instead.
The request field type of the POST /servers/{id}/actions/enable_rescue endpoint no longer accepts the value linux32.
For security reasons, the Hetzner Cloud Firewall no longer inspects protocols like FTP, SIP and PPTP to implicitly open other, related ports.
If your server has a Cloud Firewall attached and you use these protocols in a way that requires additional related ports to be accepted, then you should add a Firewall rule for the required port ranges.
See the Firewalls FAQ for details.
The Hetzner Cloud packer plugin
moved from github.com/hashicorp/packer-plugin-hcloud to
github.com/hetznercloud/packer-plugin-hcloud.
Make sure to update the plugin source in your Packer configuration:
packer {
required_plugins {
hcloud = {
- source = "github.com/hashicorp/hcloud"
+ source = "github.com/hetznercloud/hcloud"
version = ">= 1.2.0"
}
}
}
See the plugin documentation for more details.
To better communicate deprecations to API Users, we added the new response field deprecation to all endpoints that return "ISOs".
The field describes if, when, and how the resource was deprecated. If this field is set to null, the resource is not deprecated. If it has a value, it is considered deprecated.
Learn more about the field in our API docs.
This new field replaces the old deprecated field.
The Network feature to expose routes of a cloud network to a vSwitch now supports the default route (0.0.0.0/0 as destination).
The response field deprecated of all endpoints that return "ISOs" is deprecated. Please use the new field deprecation instead.
Network subnets with type "vswitch" now have a new limitation to prevent users from producing setups that don't work.
You now need to have a subnet that is smaller than the overall ip_range of the parent Network.
Example:
| # | Network ip_range | Subnet ip_range |
|---|---|---|
| correct | 10.0.0.0/16 | 10.0.1.0/24 |
| incorrect | 10.0.0.0/16 | 10.0.0.0/16 |
The old server plans with dedicated AMD vCPUs (CCX12, CCX22, CCX32, CCX42, CCX52, CCX62) are no longer available for order. Existing servers are not affected by this deprecation and they will continue to work. If you want to migrate an existing server to one of the new server plans with AMD vCPUs, you can use the rescale option.
These server types are not available through the API List endpoint. This also affects any usage "by name".
Applies to server types:
33)34)35)36)37)38)Our API reference documentation previously stated that the action list endpoints GET /v1/actions and
GET /v1/<resource>/actions accept the values progress, progress:asc and progress:desc for the query
parameter sort.
If used in a request, this caused the request to fail. We have removed the values for the sort parameter from our API docs.
Starting on 27 September 2023, the old server plans with dedicated AMD vCPUs will no longer be available for order. Existing servers are not affected by this deprecation and they will continue to work. If you want to migrate an existing server to one of the new server plans with AMD vCPUs, you can use the rescale option.
These server types will not be available through the API List endpoint after the announced date. This also affects any usage "by name".
Applies to server types:
33)34)35)36)37)38)Six new server types with dedicated AMD vCPUs are now available:
Learn more about these instances in this news article
Please note that these and new upcoming server types will use UEFI by default. If a server gets created from a snapshot, please make sure that the snapshot has an ESP / is UEFI-compatible.
Update 2025-01-30: The Actions List endpoint is not fully deprecated anymore.
GET /actions is deprecated. Starting 1 October 2023, it will no longer be available.
For more information about alternatives, please refer to the previous announcement "Resource Action endpoints" from 29 June 2023.
Also starting on 1 October 2023, all action endpoints will only return actions from the past 90 days.
All server plans with Intel® vCPUs (CCX11, CCX21, CCX31, CCX41, CCX51) are no longer available for order. Existing servers are not affected by this deprecation and they will continue to work. If you want to migrate an existing server with Intel® vCPUs to a server plan with AMD vCPUs, you can use the rescale option.
These server types are not available through the API List endpoint. This also affects any usage "by name".
Applies to server types:
11)12)13)14)15)We added the new endpoints /v1/<resource>/actions and /v1/<resource>/actions/{action_id} to all resources that use actions.
The new endpoints can be used instead of the global /v1/actions endpoint, to get actions of a specific product.
You can now expose routes from a Network to the vSwitch connection.
API Changes:
expose_routes_to_vswitch was added to all endpoints that return "Networks".expose_routes_to_vswitch was added to the POST /networks endpoint.expose_routes_to_vswitch was added to the PUT /networks/{id} endpoint.Starting on 18 July 2023, all server plans with Intel® vCPUs will no longer be available for order. Existing servers are not affected by this deprecation and they will continue to work. If you want to migrate an existing server with Intel® vCPUs to a server plan with AMD vCPUs, you can use the rescale option.
These server types will not be available through the API List endpoint after the announced date. This also affects any usage "by name".
Applies to server types:
11)12)13)14)15)To better communicate deprecations to API Users, we added the new response field deprecation to all endpoints that return "Server Types".
The field describes if, when, and how the resource was deprecated. If this field is set to null, the resource is not deprecated. If it has a value, it is considered deprecated.
Learn more about the field in our API docs.
This new field replaces the old deprecated field, which contained a boolean.
The response field deprecated of all endpoints that return "Server Types" is deprecated. Please use the new field deprecation instead.
Please note that there will be a change regarding the id fields in our API. At the moment, the id fields are only 32 bit wide. However, this is no longer sufficient. As a result, we will use larger IDs in future. This change will not affect IDs of existing resources.
Starting on 1 September 2023, the first IDs will use 52-bit integers.
Four new server types with Arm64 architecture are now available:
Learn more about these instances in this news article
Our API had an inherent assumption that all servers, images, and ISOs used the same x86 architecture. To support the launch of our Arm64 Server Types, we added new APIs to differentiate between the architectures.
The architecture fields currently support two values: x86 and arm.
architecture was added to all endpoints that return "Server Types".Previously, the image name uniquely identified the image. This is no longer true. Instead, the image is now uniquely identified through the combination of name & architecture. This means that we now have (as an example) two images named ubuntu-22.04, one with architecture x86 and one arm. If you create servers from images (or snapshots) by passing the ID, you need to take care that you pass an image with a compatible architecture. If you create servers by passing the image name, our API will select the correct image for you.
architecture was added to all endpoints that return "Images".architecture was added to the GET /images endpoint.architecture was added to all endpoints that return "ISOs". This field can be null if the ISO is not restricted to a single architecture. We do not set the architecture for ISOs uploaded on user request.architecture was added to the GET /isos endpoint.include_architecture_wildcard was added to the GET /isos endpoint.AlmaLinux 8 and 9 are now available as images for your servers. They have the image names alma-8 & alma-9.
Three new Apps are available for your servers:
owncast)photoprism)rustdesk)The new location Hillsboro, Oregon (hil) is now available for all products. Some server types might not be available in the new location.
The datacenter hil-dc1 was also added to the API and belongs to the hil location.
The network zone us-west was also added to the API. It currently includes the hil location.
Four new Apps are available for your servers:
go)ruby)prometheus-grafana)collab-tools)The app BigBlueButton (big-blue-button) is no longer available for new servers.
When you create a cloud server, it will no longer automatically include public IP addresses (IPv4, IPv6). Instead, you can now decide for yourself which types of IPs the server should include or not. And even after creating the server, you can still change your network option by adding, removing, or swapping the server’s Primary IPs. Learn more in this news article.
All endpoints under the "Primary IPs" & "Primary IP Actions" sections were added.
GET /primary_ips was added.POST /primary_ips was added.DELETE /primary_ips/{id} was added.GET /primary_ips/{id} was added.PUT /primary_ips/{id} was added.POST /primary_ips/{id}/actions/assign was added.POST /primary_ips/{id}/actions/change_dns_ptr was added.POST /primary_ips/{id}/actions/change_protection was added.POST /primary_ips/{id}/actions/unassign was added.server.public_net.ipv4 and server.public_net.ipv6 are now nullable in all endpoints that return "Servers".public_net was added to the POST /servers endpoint.