fix(test): wait for iperf server migration before continuity checks#2649
Open
nevermarine wants to merge 32 commits into
Open
fix(test): wait for iperf server migration before continuity checks#2649nevermarine wants to merge 32 commits into
nevermarine wants to merge 32 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>
Description Removes CDI (Containerized Data Importer) from the platform. Disk and image provisioning no longer goes through CDI DataVolumes — the module now owns the whole import pipeline: The VD/VI controllers create and track importer/uploader pods and target PVCs directly (new source/step pipeline instead of the DataVolume path). A new PVC populator controller populates target PVCs from DVCR; host-assigned PVC cloning uses nbdcopy; the clone strategy (snapshot vs host-assigned) is picked from the StorageProfile, cross-CSI PVC-to-PVC cloning is forbidden. The CDI importer sources are slimmed down into a pvc-importer image; import progress is streamed live from the target pod into the resource status. CDI deployments (cdi-operator, cdi-apiserver, cdi-controller, cdi-cloner) and their templates/RBAC are removed; a module hook cleans up CDI resources on upgrade while keeping the CDI config CRD for safe rollback to main. e2e: expanded block-device coverage with progress/quota observers, WFFC and cross-CSI scenarios. Bugs fixed along the way Progress & phases Import progress could jump backwards after an importer pod restart (metric reset to 0); it is now monotonic (c2a61a4). An optimistic "0%" surfaced as an intermediate Pending 0% status; Pending now reports no progress and the Pending→Provisioning transition is atomic (d122212). Progress flip-flopped between 50%/50.0%; percent formatting is now consistent (00b93a0). VirtualImage-on-PVC never streamed intermediate progress (hardcoded 0%→50%→100%) and under-reported the produced image size, breaking downstream PVC sizing; both now come from the importer pod (680608b). A busy importer /metrics endpoint could freeze Status.Progress for the whole scrape budget (d122212). Import pipeline Interrupted PVC imports were not resumed for already-created targets (b95b230). Importer pods for VI/CVI sourced from a VI-on-PVC always mounted the source as a block device, failing with FailedMount on volumeMode: Filesystem PVCs (b79920c). VirtualImage PVC auto-sizing tripped a false "insufficient size" check (7edb744); status.format was copied from source metadata instead of the actual PVC storage type (0fa1bcf). VirtualImage WFFC imports hung in the populator flow (0fb3661); smart-clone image PVCs were never bound (8cc0139). Clone snapshots set a VolumeSnapshotClassName picked by provisioner match, which the d8-snapshot-controller webhook could reject (d59b1e2); replicated PVCs now use the snapshot clone strategy (c2f69e1). pvc-importer inherited generic provisioning pod limits, throttling imports ~30% and risking OOMKill on huge images; limits raised to CDI parity (b50d7b2). Importer/uploader/bounder pods are labeled to be excluded from project resource quotas (d122212, and quota guards in 9b7fc55). A create rejection in a terminating namespace failed the reconcile instead of being surfaced on the Ready condition (c2a61a4). Migrations & VMOP The inbound migration limiter leaked slots of VMIs deleted mid-migration, eventually freezing all migrations cluster-wide (43e4f3c). Affinity changes of VMs with local (RWO) disks never triggered a placement migration: the handler required live-migratability and ignored StorageLiveMigratable (67c29ba). The workload updater kept creating evict VMOPs that the webhook was guaranteed to deny, flooding controller logs (f6b0ac0). Why do we need it, and what problem does it solve? CDI was a heavyweight third-party dependency (4 deployments, its own CRDs and RBAC) used only as a PVC provisioning backend. Owning the import pipeline removes the DataVolume indirection, gives accurate live import progress, allows to fix bugs directly, and cuts the CVE/maintenance surface of the vendored CDI codebase. What is the expected result? Creating VirtualDisks/VirtualImages from all sources (HTTP, containerRegistry, ObjectRef, upload, blank) works as before without any CDI components in the cluster; import progress is reported in the resource status. Upgrading from main cleans up CDI resources automatically; rollback to main remains safe. --------- Signed-off-by: Dmitry Rakitin <dmitry.rakitin@flant.com>
Bump the size of the VirtualDisk restored from a VirtualImage in the VirtualImageCreation e2e suite from 400Mi to 450Mi. The change is in the shared runVirtualMachineFromImageDisk helper (test/e2e/blockdevice/virtual_image_creation.go), so it covers every context that boots a VM from a restored image disk (from a VirtualDisk, from a VirtualDiskSnapshot, from a VirtualImage on DVCR/PVC). --- Signed-off-by: Dmitry Rakitin <dmitry.rakitin@flant.com>
Integrates the new bpfbridge network binding plugin (kubevirt side: deckhouse/3p-kubevirt#99) into the virtualization module and switches VM network interfaces to it, replacing the classic tap + veth + Linux-bridge datapath used for VM networks. Signed-off-by: Dmitry Lopatin <dmitry.lopatin@flant.com> Signed-off-by: Maksim Garmonov <maksim.garmonov@flant.com>
Signed-off-by: suselz <suselz@mail.ru>
Developers debugging the module with Delve had a broken helper. The dlv build/push flow shipped a single hardcoded kubectl patch snippet that assumed every component (virtualization-controller, virtualization-api, virtualization-audit) has the same sidecars (proxy, kube-rbac-proxy). Only the controller actually has them, so applying the printed patch to the api or audit deployment failed — it referenced containers that don't exist there. --------- Signed-off-by: Daniil Antoshin <daniil.antoshin@flant.com>
…stuck uploads (#2610) Uploading an image or disk (a VirtualImage, ClusterVirtualImage, or VirtualDisk with an Upload data source) could hang forever in Pending with an empty status and no upload URL, so the image simply could not be uploaded. It happened when the shared upload host started serving an invalid TLS certificate — for example after another, broken upload resource left the host serving a default self-signed certificate. In that state every upload on the cluster was blocked, not just the broken one, and recovery required manually finding and deleting the offending resource. The controller now restores the upload host certificate on its own during reconcile, so affected uploads recover automatically, and a not-yet-ready upload endpoint no longer leaves the resource with an empty status. This PR also restores a disk-specific regression: a recent refactor of the VirtualDisk upload flow dropped the upload-host self-healing for disks, so after the upload host changed (for example when publicDomainTemplate was updated) an in-flight disk upload could get stuck. The self-healing is reinstated for disks, matching how images already behave. Signed-off-by: Pavel Tishkov <pavel.tishkov@flant.com>
New conntrack sync metrics: d8_virtualization_virtualmachine__conntrack_sync_exports_total, d8_virtualization_virtualmachine__conntrack_sync_exported_bytes_total, d8_virtualization_virtualmachine__conntrack_sync_imports_total, d8_virtualization_virtualmachine__conntrack_sync_hook_waits_total --------- Signed-off-by: Daniil Loktev <lokt.daniil@gmail.com> Signed-off-by: Daniil Loktev <70405899+loktev-d@users.noreply.github.com>
Signed-off-by: Daniil Antoshin <daniil.antoshin@flant.com>
* chore(api): integrate crd-enricher into CRD codegen and add CRD examples Signed-off-by: Vladislav Panfilov <vladislav.panfilov@flant.com>
…#2624) Description Virtual machines created without paravirtualization keep all their disks and CD-ROMs on the SATA bus forever, even after enableParavirtualization is turned on and the VM is restarted. The recorded bus of an existing device always took priority over the bus implied by the paravirtualization setting. Now, when the paravirtualization mode changes, devices move to the bus of the new mode on the restart this change already requires. Devices hot-plugged on a non-default bus (for example USB) keep their bus as before. --------- Signed-off-by: Nikita Korolev <nikita.korolev@flant.com>
… resources with d8- (#2632) * fix(observability): default audit TLS cert paths to certificates dir The mounted secret is at /etc/virtualization-audit/certificates (plural), but the flag defaults pointed at /etc/virtualization-audit/certificate, so the server failed to load TLS when started without explicit flags. Signed-off-by: Daniil Antoshin <daniil.antoshin@flant.com> * refactor(observability): prefix audit log-shipper resources with d8- and drop to- Rename the ClusterLoggingConfig and ClusterLogDestination objects created for audit shipping: to-virtualization-audit -> d8-virtualization-audit and virtualization-audit-* -> d8-virtualization-audit-*, aligning with the d8- naming convention for module-owned cluster-scoped resources. Signed-off-by: Daniil Antoshin <daniil.antoshin@flant.com> --------- Signed-off-by: Daniil Antoshin <daniil.antoshin@flant.com>
Signed-off-by: Vladislav Panfilov <vladislav.panfilov@flant.com>
update CODEOWNERS Signed-off-by: Vladislav Panfilov <vladislav.panfilov@flant.com>
Signed-off-by: Maksim Fedotov <maksim.fedotov@flant.com>
Signed-off-by: Maksim Fedotov <maksim.fedotov@flant.com>
Signed-off-by: Valeriy Khorunzhin <valeriy.khorunzhin@flant.com>
verifyIPerfContinuityAfterUpgrade fetched the iperf-server VM immediately, but the post-upgrade workload migration is asynchronous and may start several minutes after the module becomes Ready. Status.MigrationState was still nil, so getIPerfClientReport panicked dereferencing it. Wait for the live migration to succeed before reading MigrationState, and guard the dereference so a future regression fails with a clear message. 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
deckhouse-BOaTswain
requested review from
diafour,
fl64,
goganat,
universal-itengineer,
yaroslavborbat and
z9r5
as code owners
July 16, 2026 17:28
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
13 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.
What & why
The post-upgrade release smoke test panicked with a nil-pointer dereference:
Root cause:
getIPerfClientReportdereferencesiperfServer.Status.MigrationState(a*VirtualMachineMigrationState).verifyIPerfContinuityAfterUpgradefetched the iperf-server VM immediately, but the post-upgrade workload update live-migrates the VM asynchronously — the migration can start several minutes after the module becomes Ready. At the time of the checkMigrationStatewas stillnil, causing the panic. The test never waited for the migration to happen.This assertion is test-only and was added recently (#2617); it is not a product regression.
Fix (test-only)
waitForIPerfServerMigrationintest/e2e/release/iperf.go(styled likewaitForIPerfServerToStart) that polls the iperf-server VirtualMachine viaf.Clients.VirtClient()untilStatus.MigrationState != nil,Result == MigrationResultSucceeded, and bothStartTimestamp/EndTimestampare non-zero. Usesframework.MaxTimeout/framework.PollingInterval.verifyIPerfContinuityAfterUpgradenow waits for the migration (newBy(...)step) and passes the fully-populated VM togetIPerfClientReport.getIPerfClientReportgets an explicitExpect(...MigrationState).NotTo(BeNil(), ...)guard so a future regression fails with a clear message instead of a panic.Verification
go build ./release/...andgo vet ./release/...pass.golangci-lint fmt ./release/...(gofumpt/gci/goimports) reports no changes. Note:golangci-lint runcurrently panics in this environment due to a Go toolchain vs linter version mismatch (Go 1.26 vs linter built with Go 1.25) during package type-checking — unrelated to this change.