fix(disks): restore disk from snapshot when storage rounds size up#2635
Draft
nevermarine wants to merge 16 commits into
Draft
fix(disks): restore disk from snapshot when storage rounds size up#2635nevermarine wants to merge 16 commits into
nevermarine wants to merge 16 commits into
Conversation
Signed-off-by: Dmitry Rakitin <dmitry.rakitin@flant.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Changelog v1.9.3 Signed-off-by: deckhouse-BOaTswain <89150800+deckhouse-boatswain@users.noreply.github.com> Co-authored-by: universal-itengineer <universal-itengineer@users.noreply.github.com>
Signed-off-by: Valeriy Khorunzhin <valeriy.khorunzhin@flant.com>
During a restore the VM is stopped inside the maintenance window: its KVVMI and then the KVVM itself are deleted. Bringing the VM back up afterwards relied on an implicit side effect — the recreated KVVM getting RunStrategy=Always at creation time. That path is racy and only covers the always-on policies. --------- Signed-off-by: Daniil Antoshin <daniil.antoshin@flant.com>
Signed-off-by: Maksim Fedotov <maksim.fedotov@flant.com>
Add validation to getIPerfClientReport verifying that the iperf client session spans the virtual machine migration: start time before migration start, end time after migration end, and no more than one zero-byte interval during the process. Signed-off-by: Dmitry Prytkov <dmitry.prytkov@flant.com>
…refix (#2594) The d8 v ansible-inventory command read host variables from the bare provisioning.virtualization.deckhouse.io/ prefix, while the virtualization-provisioner module reads them from the separate vars.provisioning.virtualization.deckhouse.io/ prefix. The same VM annotations produced different inventories depending on the tool. Read host variables from the vars. prefix to match the module. Groups still come from provisioning.virtualization.deckhouse.io/groups. Update the command help, usage examples and the ansible-inventory FAQ. Signed-off-by: Pavel Tishkov <pavel.tishkov@flant.com>
…d manually (#2596) fix(vm): fix VM stuck until the child KVVMI in Failed phase is deleted manually Signed-off-by: Yaroslav Borbat <yaroslav.borbat@flant.com>
Virtualization has no primitive to manage a group of identical virtual machines whose count changes over time. Every "I need N identical VMs and the number varies" scenario — CI runner fleets, VDI desktop pools — has to be solved with orchestration outside the platform: users write their own controllers/scripts to create and delete VirtualMachines, watch their number, recreate lost ones and clean up their disks. This duplicates logic and is error-prone around races and node failures. This PR introduces VirtualMachinePool (paid editions only, EE/SE+): a namespaced resource that declaratively keeps a requested number of identical VMs and integrates with kubectl scale, HPA and KEDA through the standard scale subresource. Its template is an ordinary VirtualMachineSpec, so a replica is no different from a manually created VM. The feature is complete and covers: Replica management — keeps the requested number of identical VMs, replaces lost ones, and reports state in status (replicas, readyReplicas, selector, conditions). It is cache-lag-safe (ReplicaSet-style expectations), so a lagging informer cache cannot double-create replicas. Scale-down policy — NewestFirst / OldestFirst choose which replica leaves on anonymous scale-down via scale; Explicit forbids anonymous shrink through a webhook, so for "busy" workloads (CI runners, VDI) replicas can only be removed by address. Addressed removal — the scaleDownWith subresource deletes named replicas and shrinks the pool by that count, instead of letting the controller pick victims. In-place template propagation — editing the template rolls the change out to existing replicas; disruptive changes wait for a restart and are surfaced via status.restartPendingReplicas. Reusable disks — per-replica disks are described in virtualDiskTemplates with a reclaim policy: Delete disks belong to the VM and are removed with it; Retain disks belong to the pool, outlive the replica and are reused on the next scale-up, with an optional warm buffer (keep) and TTL garbage-collection. The resource is available only in paid editions, gated behind the VirtualMachinePool module feature gate (default off, locked off in CE). The API/CRD installs in every edition, but the controller self-gates on the feature gate, so the resource does nothing in CE. --------- Signed-off-by: Pavel Tishkov <pavel.tishkov@flant.com> Signed-off-by: Vladislav Panfilov <vladislav.panfilov@flant.com> Co-authored-by: Vladislav Panfilov <vladislav.panfilov@flant.com>
…ate (#2621) The VirtualMachinePool feature gate was an opt-in placeholder. As the feature has not been released yet, drop the gate before release so the resource is available in EE/SE+ out of the box and stays unavailable in CE — the same model as VolumeMigration/TargetMigration. Nothing to migrate: the gate never shipped. - gate becomes locked-on in EE/SE+, locked-off in CE (no user opt-in) - dropped from the user-facing featureGates enum - pool webhooks gated by edition in helm instead of the feature-gate list Controller/webhooks keep self-gating on the now-locked gate. Signed-off-by: Pavel Tishkov <pavel.tishkov@flant.com>
Adds: - liveMigration.systemNetworkName ModuleConfig field. When set, VM live-migration traffic is routed over the per-node interfaces of a SystemNetwork CR from the SDN instead of the default node network. - migrationiface controller resolves the per-node kernel interface name and writes it onto each Node as the annotation virtualization.deckhouse.io/migration-iface. virt-handler reads that annotation at startup and binds its migration proxy to the SystemNetwork's per-node IP. - ModuleConfig validator to check if SystemNetwork added to liveMigration.systemNetworkName exists. - New reason for VMOP if migration network is unavailable on nodes. --------- Signed-off-by: Daniil Loktev <lokt.daniil@gmail.com> Signed-off-by: Daniil Loktev <70405899+loktev-d@users.noreply.github.com> Signed-off-by: Vladislav Panfilov <97229646+prismagod@users.noreply.github.com> Co-authored-by: Vladislav Panfilov <97229646+prismagod@users.noreply.github.com>
Description Log in to the private RU registry in the release channels workflow and check module versions against registry.deckhouse.ru/deckhouse. Why do we need it, and what problem does it solve? Release channel checks need access to artifacts in the RU registry. This keeps the workflow and version check pointed at the registry that contains the expected release artifacts. --------- Signed-off-by: Nikita Korolev <nikita.korolev@flant.com>
…lose gaps (#2602) Sizing policy validation errors were confusing and, in some cases, missing. When a VirtualMachineClass sets per-core memory limits, the error reported the allowed values per CPU core, while users configure the total memory — so the suggested numbers matched nothing they could actually set. Messages now report the total memory to set for the current number of cores, name the field to change, and combine multiple violations into a single readable list. This also fixes cases where the policy was silently not enforced: the CPU cores step was never validated, and a memory / per-core minimum was ignored when no maximum was set. Signed-off-by: Pavel Tishkov <pavel.tishkov@flant.com>
* docs: add VM launcher update notes to release notes Signed-off-by: Vladislav Panfilov <vladislav.panfilov@flant.com>
When a VirtualDisk is created without an explicit size, its PVC is requested at the data source's virtual size. Storage backends that round volume sizes up (e.g. Ceph RBD rounds to whole GiB) provision a larger volume, so the VolumeSnapshot restore size exceeds the originally requested size stored in the snapshot metadata. On restore that stored original size was validated against the snapshot restore size and rejected as insufficient, leaving the restored disk Failed and the VirtualMachineOperation restore stuck in Failed. Use the restore size as a lower bound when the size comes from the snapshot metadata rather than an explicit user request. Signed-off-by: Maksim Fedotov <maksim.fedotov@flant.com>
deckhouse-BOaTswain
force-pushed
the
main
branch
from
July 16, 2026 17:28
ecd65ac to
a9c8af4
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
11 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Restoring a virtual disk (and therefore a virtual machine) from a snapshot fails on storage backends that round volume sizes up to a fixed granularity — for example Ceph RBD rounds up to whole GiB. The restored disk stays in the
Failedphase with aProvisioningFailed/ "the specified pvc size is insufficient" error, and anyVirtualMachineOperationrestore that depends on it never reachesCompleted.It happens when the source disk was created without an explicit size: its volume is requested at the source image's virtual size, but the backend provisions a larger volume, so the snapshot's restore size ends up bigger than the size recorded in the snapshot metadata. On restore that smaller recorded size was rejected as insufficient. It is now treated as a lower bound and raised to the snapshot restore size, but only when the size was not explicitly set by the user (an explicit too-small user request is still rejected).
Why do we need it, and what problem does it solve?
On rounding storage backends (Ceph RBD is the common case) snapshot restore is effectively broken for any disk that was provisioned without an explicit size — the restored disk and the restore operation are stuck in
Failed. This restores the ability to recover disks and VMs from snapshots on such storage.What is the expected result?
Reproduction (on a Ceph RBD storage class):
VirtualDiskfrom aClusterVirtualImagewithspec.persistentVolumeClaim: {}(no size). The PVC is requested at the image virtual size but provisioned rounded up to the next GiB.VirtualDiskSnapshotof it.VirtualDiskfrom that snapshot with no explicit size.Before: the new disk goes to
Failed(the specified pvc size is insufficient).After: the new disk provisions at the snapshot restore size and becomes
Ready; VMOP restore completes.Checklist
Changelog entries