Skip to main content

Changelog

Reviewed on · View on GitHub

Discover the latest changes to Numspot products and features.
Last updated on 19 October 2026
October 2026
PostgreSQL
Changed
Connection behaviour and timeouts documented
A new PostgreSQL guide documents the connection behaviour of managed clusters: the 60-second idle connection timeout applied by the connection endpoint (idle sessions, long silent operations such as CREATE INDEX, VACUUM or pg_restore are cut, while client-side TCP keepalives do not prevent the disconnect), the fact that a cut statement keeps running server-side, and a consolidated view of the connection limits (PgBouncer pooling limits, max_connections) with troubleshooting guidance.
Changed
Delete a PostgreSQL cluster page aligned with the console
The Delete a PostgreSQL cluster action page now reflects the current console: the confirmation dialog ("Delete cluster") asks to type the cluster name in the "Name of cluster" field and confirms with "Delete permanently", and on the cluster details page the delete action lives in the ⋮ menu. A step-by-step console walkthrough was added to the English page and all console screenshots were refreshed.
Changed
PostgreSQL connection limits documented
The auto-tuning page now documents the connection limits of managed PostgreSQL clusters: max_connections fixed at 100 per instance with 3 slots reserved for the platform, all client connections going through the built-in PgBouncer pooler (3 replicas, session mode), up to 300 concurrent client connections (max_client_conn of 100 per replica) and 60 simultaneously active sessions (default_pool_size of 20 per database and user pool), with idle client connections closed after 10 minutes.
Added
PostgreSQL cluster restore
Added a page documenting how to restore a PostgreSQL cluster from a backup or from the WAL archive of a source cluster, with point-in-time recovery (PITR) through the pointInTime parameter. The restore creates a new cluster; the source cluster is not modified.
IAM
Changed
Permissions explorer: tenants, UUIDs and standard roles
The IAM permissions reference now covers the full permission catalog (142 permissions), each with its tenant (organisation, space or both) and its UUID. It also adds the standard roles catalog (organisation and space levels) with the permissions each role grants, filterable by role; permissions with no public API, performed from the Numspot console, are listed and flagged.
Added
Service account access keys (AK/SK)
New pages documenting the management of durable AK/SK access key pairs attached to a service account: creation (at most two active pairs per account, lifetime from 1 hour to 1 year, secret key displayed only once), list and details with the access key masked, and deletion which also revokes the sessions opened with the deleted key. These credentials authenticate programmatic access to the Object Storage.
Container Registry
Changed
Default administrator username documented
The Container Registry pages now surface the default administrator account: the username is fixed to admin and cannot be changed, only the password is generated and retrieved through the dedicated action, and resetting the password does not change the username.
Added
Large image push limits and troubleshooting
New guides documenting the limits for pushing large images to the Container Registry: ingress proxy read/send timeouts raised to 3600 s, observed push throughput of ~4–5 MB/s, recommended blob (layer) size below ~300 MB, plus best practices for pushing large images (standard streaming clients, small layers, .dockerignore) and troubleshooting for the "unexpected EOF" error.
Compute
Changed
Boot script (user data) documented on the VM creation page
The "Create a VM" page now documents the boot script option ("user data"): the UserData parameter of the creation request (Base64-encoded, up to 500 KiB), the Windows wrappers (PowerShell, VBScript), the Linux cloud-init #cloud-config format and the metadata endpoint that exposes the user data to the VM. User data is processed only at creation.
Resources
Removed
Resource locations page removed
The "Resource locations" page (zones, regions and endpoints) is removed from the documentation. The /resources URL keeps answering as a noindex redirect to the Catalog concepts page, which describes the Numspot regions.
September 2026
IAM
Added
Permissions and related APIs reference page
New reference page in the IAM menu: for each permission of the IAM domain, the list of APIs using it, with search and autocompletion by permission or by API (method, path or summary), live filtering and a link to each endpoint definition in the OpenAPI reference. Only the endpoints exposed on the public API are listed.
Changed
IAM documentation updated and illustrations regenerated
The 32 IAM pages are aligned with the current console and API: user and service account creation flows with roles and permissions assigned at creation time (Spaces, Roles, Permissions wizard) or without access rights, identity duplication from an existing profile, user deletion which automatically revokes their API keys, asynchronous service account deletion, service account name format and token duration bounds (1 second to 48 hours), space deletion handled through a support request, corrected activation link generation endpoint (POST /iam/organisations/{organisationId}/users/{userId}/recover), the Roles page redesigned (roles are viewed from the identity details pages, the dedicated console page no longer exists) and interactive walkthroughs regenerated with the current console screens.
Kubernetes
Added
Documentation of the Kubernetes API pages listing clusters and node profiles
Two new Kubernetes action pages: "List Kubernetes clusters" (GET /kubernetes/spaces/{spaceId}/clusters, with pagination via the nextToken field) and "List available node profiles" (GET /kubernetes/nodeProfiles).
Added
Guide for connecting a VM to a managed Kubernetes cluster
A new guide is added to the managed Kubernetes service: "Connect a VM to a managed Kubernetes cluster". It documents the private connection of a VM in a VPC to the API server of a managed cluster through a Hybrid Bridge: bridge creation, route table update with VPC peering targets, outbound flow opening towards the control plane, kubeconfig file deployment on the VM, SNI constraint on the API server URL and troubleshooting.
Added
Internal exposure guide through a private load balancer
A new guide is added to the managed Kubernetes service: "Expose an application internally through a private load balancer". It documents the creation of an internal LoadBalancer type Kubernetes Service with the annotations of the cloud controller manager of the platform: private load balancer in a subnet, filtering of the allowed sources, security groups, IP address lifecycle (auto-allocated private IPs or pre-allocated static IP), access verification from a peered VPC and troubleshooting.
PostgreSQL
Added
Documentation of the PostgreSQL API pages listing versions, clusters and backups
Four new PostgreSQL action pages: "List available PostgreSQL versions" (GET /postgresql/versions), "List PostgreSQL clusters" (GET /postgresql/spaces/{spaceId}/clusters), "List the backups of a PostgreSQL space" (GET /postgresql/spaces/{spaceId}/backups) and "Get a PostgreSQL backup" (GET /postgresql/spaces/{spaceId}/clusters/{clusterId}/backups/{backupId}).
Changed
Details of available extensions and license restrictions
The PostgreSQL cluster creation page now details the TimescaleDB and Vector extensions, their use cases, the Apache 2 license restriction for TimescaleDB, the feature differences between Apache 2 and Community editions, and the fact that an extension must be selected at cluster creation.
Container Registry
Added
Documentation of the Container Registry API pages listing registries and resetting the administrator password
Two new Container Registry action pages: "List registries" (GET /registry/spaces/{spaceId}/registries) and "Reset the administrator password" (PUT /registry/spaces/{spaceId}/registries/{registryId}/password/reset, 202 Accepted response with an empty body, new password to be retrieved via the password read request).
Fixed
Correction of the Harbor registry host in the response examples
The response examples of the creation page now use the real format of the Harbor registry host (reg-<id>.space-<id>.hcp.cloudgouv-eu-west-1.numspot.com) instead of a generic host.
Secret Manager
Added
Documentation of retrieving the admin token of a Secret Manager instance
New Secret Manager action page: "Get the admin token of a Secret Manager instance" (GET /secrets/spaces/{spaceId}/instances/{instanceId}/password).
Console
Added
First steps with Numspot page
New "First steps with Numspot" page describing the first user's journey in the console, from the least privilege principle to the first resources: rights verification, first space creation, user creation and activation link generation, assignment of roles, permissions and spaces, then creation of a first VM and the first managed services (Kubernetes, PostgreSQL). A note recalls the arrival journey: account opening request, first super admin user with a recommended generic email address, and 48-hour account activation.
Network
Changed
Documentation of automatic public IP assignment via mapPublicIpOnLaunch
The mapPublicIpOnLaunch attribute of subnets is now documented: enabling it through the subnet attribute update, automatic assignment of a public IP (chosen randomly) to the primary NIC of each VM launched in the subnet, persistence of the link across stops and starts, deletion of the IP with the VM, absence from the "Public IP" section, and Internet access requirements (internet gateway, route table, security group).
Connectivity
Changed
IPsec securing of VPN tunnels described accurately
The VPN connection concepts page corrects the description of how the tunnel is secured: IPsec is not an encryption protocol but a suite of protocols, with IKEv2 negotiating the keys and ESP encrypting the data. A link to the VPN Configuration Reference page is added. This reference page clarifies the role of each phase (phase 1: IKE SA, phase 2: ESP child SA), explains the HMAC authentication required by non-AEAD algorithms and adds the mandatory lifetimes (28800 seconds for phase 1, 3600 seconds for phase 2).
Changed
Hybrid Bridge pages aligned on the API and Terraform
The Hybrid Bridge action pages (create, read, delete) are aligned with the Compute Bridges model: the Console tab is removed since Hybrid Bridges are managed through the API only, the API tab becomes the default tab and a Terraform tab is added with the link to the numspot_hybrid_bridge resource of the registry. The concepts page adds a link to the guide for connecting a VM to a managed Kubernetes cluster.
Changed
Compute Bridge: API management only
The "Compute Bridges" section is no longer available in the console. The Compute Bridge pages (creation, viewing, deletion) now document API and Terraform management only.
Object Storage
Added
Bucket deletion page published
A new "Delete a bucket" action page is added as a mirror of the creation page. It documents the deletion of an Object Storage bucket through the S3 aws s3 rb command (AWS CLI) and through Terraform (numspot_bucket resource). Deletion requires the objectstorage.bucket.delete permission and an empty bucket, in accordance with the AWS S3 standard: deleting a bucket does not delete the objects it contains, they must be deleted beforehand (or through the --force option of AWS CLI). Deleting a bucket or an object is irreversible.
Resources
Changed
Resource locations page rewritten around zones and regions
The "About Numspot Regions, Endpoints and Sub-regions" page becomes "About Numspot Zones, Regions and Endpoints" and describes the actual Numspot hierarchy: two zones, STD (standard) and SNC (SecNumCloud), currently hosting one region each, eu-west-2 and cloudgouv-eu-west-1 respectively. The "sub-regions" model (availability zones), with no reality at Numspot, is removed along with the associated diagram. The API request example URL is fixed to api.eu-west-2.numspot.com.
Removed
Quotas page removed
The Quotas page described viewing quotas from the Numspot console (the "Quotas" tab of the account) and a list of per-resource limits that do not match the current console. This page is removed; the /docs/resources/quotas URL remains alive as a noindex redirect to the Resource locations page.
Compute
Changed
Server certificates: section moved to Compute and concepts page published
The Server Certificates section is moved from IAM to Compute (new URLs /docs/compute/server-certificate/…, the old /docs/iam/server-certificate/… URLs remain alive as noindex redirects). The quick start page, which duplicated the content of the action pages, is replaced by a concepts page describing server certificates: definition, expected content (x509 certificate in PEM format, RSA PKCS1 private key, intermediate authorities chain), identification (name, path, NRN), returned metadata and use for the SSL termination of a load balancer. The creation page, previously titled "Télécharger un server certificate" (Download a server certificate) by mistake, is renamed "Créer un server certificate" (Create a server certificate) with an introduction describing the import of the certificate and its private key. The Terraform tabs of the four action pages, previously empty, now link to the documentation of the numspot_server_certificate resource.
Changed
Explicit error when starting a VM without a system volume
When starting a VM whose root device (/dev/sda1 by default) no longer has a volume attached, the API now returns a 409 Conflict error detailing the cause (missing system volume) instead of a generic "Resource conflict." message. The VM lifecycle page documents this case and its resolutions.
Changed
Resource tags: pages published and permissions clarified
The resource tag management pages (viewing, adding, deleting) are now published. The associated IAM permissions are clarified: adding or deleting tags requires the update permission of the resource in question (e.g. compute.vm.update for a VM, compute.loadBalancer.update for a load balancer), and viewing tags requires the read permission of each resource (only the tags of resources you are allowed to read are returned).
DLP
Changed
"Data Leak Protection" renamed to "Data Loss Prevention"
The DLP service is now called "Data Loss Prevention" throughout the documentation: menus, pages, glossary and API reference. The DLP acronym remains unchanged.
August 2026
DLP
Added
DLP Troubleshooting page and API authentication documentation
Added the DLP "Troubleshooting" page and the API token (Bearer) authentication section. DLP terms are merged into the glossary.
Added
DLP OpenAPI reference and walkthrough fixes
Added the DLP OpenAPI reference to the site menu and fixed the interactive screenshots (walkthroughs) of the DLP action pages.
Added
DLP managed service documentation
Added the complete DLP (Data Loss Prevention) service documentation: concepts, actions (rules, policies, detectors, publishing, logs, audit, dashboard), guides (quick start, monitoring, DSL conditions), versions and lifecycle, reversibility.
Added
Additional DLP action pages
Added 8 action pages: update/delete a rule, update/delete a policy, update/delete a custom detector, delete a user, reset a user password.
Compute
Changed
Official images update - 2026-08-10
Added the 2026-08-10 official images (AlmaLinux 9/10, Debian 12/13, RockyLinux 9/10, Ubuntu 22.04/24.04/26.04) for the eu-west-2 and cloudgouv-eu-west-1 regions. The 2026-06-17 images move to supported.
Changed
Granular IAM permissions for the compute domain
Replacement of the catch-all compute.all permission with unitary permissions (compute.vm.create, compute.vpc.get, compute.loadBalancer.update, etc.) across all action pages. Server-certificate permissions (compute.serverCertificate.*). Tags remain under compute.all. Removal of the VPC Access Point pages (not exposed).
July 2026
Mistral AI Platform
Added
Mistral AI Platform reversibility
Added the reversibility page for the Mistral AI Platform managed service.
June 2026
PostgreSQL
Changed
TLS verify-full configuration for PostgreSQL
Added, in the PostgreSQL security guide, the CA certificate chain to place in the ca.cert file for the standard and SecNumCloud zones, along with the use of the sslrootcert parameter to enable sslmode=verify-full.
Changed
PostgreSQL Console tabs redesign
Complete redesign of the Console tabs on the PostgreSQL action pages, with updated instructions and new screenshots.
Added
PostgreSQL version policy and lifecycle
Added the supported-versions policy, lifecycle and ANSSI hardening recommendations for PostgreSQL clusters.
Managed Services
Changed
Full access audit
Extended the PostgreSQL user and role management guide with a complete set of audit queries for roles, permissions, default privileges, object owners, RLS policies and active connections. Extended the Kubernetes user and role management guide with complete RBAC audit commands: roles, ClusterRoles, RoleBindings, ClusterRoleBindings, ServiceAccounts, subjects, effective permissions and sensitive access. Extended the Harbor user and role management guide with complete API commands to audit users, system administrators, projects, public projects, members, robot accounts and permissions.
Changed
Managed services documentation reorganization
Reorganization of the managed services documentation (Kubernetes, PostgreSQL, Container Registry): separation of actions (API/Console/Terraform) and guides, addition of a Reversibility page for each service, and application of the mandatory menu structure.
Support
Added
Reversibility policy
Publication of the general reversibility policy and per-service reversibility pages (Kubernetes, PostgreSQL, Container Registry) in the Support section.
Secret Manager
Added
Secret Manager documentation
Publication of the Secret Manager documentation: concepts, instance creation and management, SSO configuration and user-management guides, and reversibility.
Compute
Changed
Keypairs renamed to Key pairs
The Keypairs menu and terminology have been renamed to Key pairs to align with the official terminology.
Changed
Image catalog redesign
The image catalog is now limited to official open-source images (Debian, Ubuntu, AlmaLinux, RockyLinux) and the space's private images. The public-image concept has been removed. Sharing private images through permissions is replaced by the export/import feature via Object Storage.
Object Storage
Added
Object Storage API reference
Added the Scalar API reference page for Object Storage, documenting all available endpoints and operations.
May 2026
Container Registry
Added
Container Registry SSO/OIDC configuration
Added the SSO/OIDC configuration guide for the Container Registry, including UAA and LDAP integration and troubleshooting sections.
Added
Container Registry guides
Added the Container Registry guides: Harbor replication, proxy cache, robot accounts and vulnerability scanning.
Kubernetes
Added
Kubernetes security and compliance guides
Added guides for installing Falco and metrics-server, configuring secure authentication and managing users and roles for Kubernetes clusters.
Added
Backup and restore with Velero
Added the guide for backing up and restoring Kubernetes clusters with Velero, covering configuration, backup and restore operations.
PostgreSQL
Added
PostgreSQL user management and security
Added documentation on user and role management, secure authentication, and account blocking and deletion procedures for PostgreSQL clusters.
Added
PostgreSQL auto-tuning
Added the PostgreSQL auto-tuning documentation: automatic optimization of cluster configuration based on instance size.
April 2026
PostgreSQL
Added
PostgreSQL backup management
New backup management feature for PostgreSQL clusters: list, create and delete on-demand backups. Automatic backups are also available.
March 2026
Connectivity
Added
Hybrid Bridge documentation
Publication of the Hybrid Bridge documentation: concepts, configuration and management of hybrid bridges to interconnect your on-premises infrastructure with Numspot.
Inventory
Added
Catalog and Inventory documentation
Publication of the Catalog and Inventory documentation: centralized exploration and management of space resources.
February 2026
Container Registry
Added
New Container Registry managed service
The Container Registry managed service lets you store, manage and secure software artifacts.
Inventory
Added
New inventory service
Our inventory service, available via API and in the console, lets you list and explore all the resources linked to your PostgreSQL clusters (pods, persistent volumes) and Kubernetes clusters (nodes, load balancers, persistent volumes). This feature provides a centralized, detailed view of all active resources in each customer space, with the ability to group them by parent-child hierarchy for simpler analysis. Ideal for auditing, optimizing or monitoring your managed environments, this tool makes real-time infrastructure management easier.
PostgreSQL
Added
PostgreSQL cluster creation from the console
Added the Console tab for creating PostgreSQL clusters, letting you create a cluster directly from the Numspot interface.
January 2026
Compute
Changed
Snapshot and image export
Added documentation for exporting snapshots and images to Object Storage, enabling data portability and reversibility.
Kubernetes
Added
New Kubernetes managed service
The Kubernetes managed service lets you create highly available Kubernetes clusters, with GPU support where required for intensive-compute workloads, scientific simulation, big-data analysis and real-time processing, offering an architecture optimized for performance, memory and operational stability at scale.
PostgreSQL
Added
New PostgreSQL managed service
The PostgreSQL managed service lets you create highly available PostgreSQL relational databases.
October 2025
Compute
Added
Security Groups documentation
Publication of the Security Groups documentation: concepts, creation and management of security groups for filtering network traffic.
September 2025
Compute
Added
New v7 VM offer
The v7 VMs are designed for intensive-compute workloads, scientific simulation, big-data analysis and real-time processing, offering an architecture optimized for performance, memory and operational stability at scale.
August 2025
Compute
Added
New Inference VM offer
Launch of a new range of virtual machines optimized for AI inference, delivering greater GPU performance, reduced latency and improved efficiency for deploying machine-learning models in production.
July 2025
Object Storage
Added
Object Storage documentation
Publication of the Object Storage documentation: concepts, creation and management of buckets and objects.
Compute
Added
Key pairs documentation
Publication of the Key pairs documentation: creation, import and management of SSH key pairs for instance access.
June 2025
Connectivity
Added
VPN documentation
Publication of the VPN documentation: configuration, management and use of VPN connections to secure access to your infrastructure.