Hikube API v1.6.4

This release keeps a resource that is being deleted visible, paginates the lists of instances and disks, and lists the subnets of a whole project in a single call. It also protects the system disk of a running VM, and refuses upfront a disk that the project's storage quota cannot hold. The additions remove nothing from the existing contract.

Behaviour change: a deleted resource no longer disappears from the API immediately. It stays readable with the status deleting until it is actually removed (see Changed).

Added

• Network: subnets of a project — GET /network/v1alpha1/projects/{project_id}/subnets returns in a single call the subnets of all the project's VPCs, instead of one call per VPC.

• Compute and disks: pagination — GET .../instances and GET .../disks accept pageSize (500 at most; a larger value is served as 500) and pageToken, the nextPageToken of the previous response. Items are sorted by name, and totalCount gives the size of the whole collection on every page. nextPageToken is empty on the last page. Without these parameters, the list is returned in full, as before.

• Error codes — They are returned with HTTP 400. The error format is described in Errors.

Code	Service	Meaning
10009	All	Invalid page request: negative pageSize, or a pageToken that does not come from this list
10010	All	The resource is being deleted: it can be read, but no longer changed, and its name cannot be reused until it is gone
20208	Compute	A VM's system disk can only be replaced, detached or moved while the VM is stopped: stop it first
20209	Compute	The VM cannot start yet: its latest change is still being applied (about 30 seconds). Retry
20210	Compute	The disks that the creation or the update adds or enlarges do not fit in the project's remaining storage quota
30202	Disks	The disk to create, or its new size, does not fit in the project's remaining storage quota

Changed

These changes modify an existing behaviour. Check them if you call the API from your own tools.

• All resources — After its deletion, a resource stays readable and listed with the status deleting, until it has actually disappeared; a GET then returns 404. In the meantime, any change to the resource or to its sub-resources, and any creation under the same name, are refused with code 10010. A new deletion request is still accepted. Before, the resource returned 404 as soon as deletion was requested, although it still existed. In real time, the console receives status_changed with deleting, then deleted.

• Compute — A VM's system disk (disks[0]) can only be replaced, detached or moved if the VM is stopped (stopped). Otherwise, the update is refused with code 20208 and nothing is changed.

• Compute — Right after an update, starting a VM is refused with code 20209 until the update has been applied, which takes about 30 seconds. Starting earlier could make the VM start on its previous configuration, for example on a system disk that had just been replaced. This refusal can also occur a few seconds after a VM is stopped: retry.

• Disks and compute — A disk to create or enlarge is checked against the project's remaining storage quota. If it does not fit, the request is refused with code 30202 (disks) or 20210 (creation or update of a VM), and nothing is created. details.storageQuota gives the requested size and the remaining space. Before, the disk was accepted, then stayed stuck for lack of space.

• Quotas — For a project that has sub-projects, the quota shown and checked is the one the project really has left, once the share given to its sub-projects is deducted. This is the quota that applies to new resources, to the size check of the Kubernetes control plane (70209) and to the console's real-time updates.

Fixed

• Compute — A stopped VM stays stopped when you update it or during a platform maintenance. It could restart on its own.

• MariaDB — Creating a user was refused with an authorization error since v1.6.3. It works again.