- Add teardown() function for reliable cleanup after each test
- Increase HelmRelease wait timeout from 5s to 30s
- Switch base disk in source.disk test from source.http to source.image
- Remove inline cleanup in favor of teardown
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Myasnikov Daniil <myasnikovdaniil2001@gmail.com>
The clone step is what this test exercises, not the HTTP import. Ubuntu
noble is ~600MB; Alpine is ~50MB, which cuts CI time without losing
coverage.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Myasnikov Daniil <myasnikovdaniil2001@gmail.com>
Follow-up to PR #2258: adds two new @test cases in hack/e2e-apps/vm-disk.bats
covering the new vm-disk clone paths — source.image (golden image from
cozy-public) and source.disk (PVC clone within tenant namespace). Existing
source.http coverage in vminstance.bats is not touched.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Myasnikov Daniil <myasnikovdaniil2001@gmail.com>
- Log .status.history regardless of content so a silently empty result
(Flux field rename) shows up in CI logs, and treat empty history on
a Ready HelmRelease as a distinct failure. A Ready HR by definition
has at least one snapshot; anything else is a shape-drift signal.
- Replace the unquoted heredoc in remediation-guard.sh with a printf |
grep pipeline. printf %s treats statuses as literal payload (no $
expansion surprises for future callers), grep --quiet --extended-regexp
returns exit status the caller can forward directly.
- Share the etcd-absent values file between both invariant tests
(packages/apps/kubernetes/tests/values-ci-no-etcd.yaml) instead of
duplicating the --set block.
- Fix typo "override applied" -> "override is applied" in the
Kubernetes ApplicationDefinition.
- Add a coupling comment in the ApplicationDefinition annotation that
points at the wait-for-kubeconfig init deadline in _helpers.tpl, so
a future operator raising the HR timeout updates the init deadline
too.
- Clarify the per-annotation timeout comment in rest.go so it stops
implying the feature is Kubernetes-only (it is not - only today's
one user is).
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Flux helm-controller's ClearFailures() zeroes installFailures and
upgradeFailures on every successful reconciliation (see the upstream
HelmReleaseStatus method). The previous guard ran after the HelmRelease
was Ready, at which point the counters were always 0 - the assertion
was vacuous and would have passed against a reverted fix.
Switch to .status.history, which retains per-revision release
Snapshots that survive a subsequent successful reconciliation. A
remediation cycle leaves behind a Snapshot with status=uninstalled
(the install-remediation code path) or status=failed (Helm release
failure that remediation then uninstalled). Either one signals the
race actually fired.
Rewrite the bats unit tests to cover: empty history, deployed-only,
deployed+superseded (happy path - not detected), single failed,
single uninstalled, uninstalled-then-deployed, and deployed-then-failed
(all detected). The pinned-shape test feeds a realistic HR status
snippet through yq the same way run-kubernetes.sh does via kubectl
-o jsonpath.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
run-kubernetes.sh extracts .status.installFailures and
.status.upgradeFailures via kubectl -o jsonpath. If a future flux
release renames the counters, kubectl returns empty, the guard reports
no cycle, and e2e silently misses real remediation loops.
Add a bats unit test that feeds a pinned HelmRelease v2 status snippet
through the same jsonpath and asserts the extraction still yields the
expected values. Also leave a pointer comment in run-kubernetes.sh so a
future flux bump surfaces the version coupling in review.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Shell && and || have equal precedence and left-to-right associativity,
so the previous guard parsed as (((A && B) || C) && D) and silently
passed on the canonical failure mode: install_failures=1 with an empty
upgrade_failures.
Extract the check into helmrelease_has_remediation_cycle() in a
dedicated helper sourced from run-kubernetes.sh, and add unit tests
under hack/remediation-guard.bats that pin the expected behavior for
every combination of empty, zero, and positive counter values.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Before cleanup, inspect the parent HelmRelease installFailures and
upgradeFailures counters. A non-zero value means flux helm-controller
hit its wait timeout, ran install/upgrade remediation (uninstall),
and re-installed - the exact race condition this PR closes. Fail the
bats test in that case so the signal surfaces in CI instead of being
masked by a green retry.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Increase Kafka CR readiness timeout from 60s to 300s to account for
slow Strimzi startup on QEMU-based CI sandbox (4 JVM pods).
Add wait-for-delete before re-applying to prevent race condition where
kubectl apply hits a still-deleting resource on retry attempts.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Root cause: EXIT trap called _k8s_test_cleanup after subshell exit,
but variables ${port} and ${test_name} were not available in that
context. set -u caught the unset variable and exited with error,
marking a successful test as failed. This caused every kubernetes
test to "fail" on attempt 1, then retry 2 more times with real
failures (NFS PVC timeout because the cluster was being deleted).
Fix: remove trap and cleanup function entirely. Pre-cleanup at
test start handles stale resources. Inline cleanup at test end
handles normal path. On failure, resources are cleaned up by
pre-cleanup on the next retry.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Root cause of NFS test failures: port-forward had `timeout 500s`
(8.3 minutes), but tests after it take up to 26 minutes (node
join 8m + node Ready 5m + LB test 2.5m + NFS 10m). Port-forward
was guaranteed to die mid-test.
Remove the timeout entirely. Cleanup is already handled by:
- EXIT trap (_k8s_test_cleanup does pkill port-forward)
- Explicit cleanup at every exit point
- Job-level timeout-minutes: 120 as backstop
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
- Increase NFS PVC Bound timeout from 2m to 5m — RWX via kubevirt
CSI provisions an NFS server pod which takes time
- Add pod describe and events dump when NFS test pod fails to
complete, making the root cause visible in CI logs
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
The explicit _k8s_test_cleanup calls cover known error paths, but
set -e can terminate the script at any kubectl/wait command. Add
an EXIT trap to catch these unexpected exits. The function runs
in a cozytest.sh subshell, so the trap is scoped and does not
affect parent traps. Removed 2>/dev/null from kubectl delete in
cleanup to preserve useful error output.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Use a _k8s_test_cleanup helper called at every exit point
instead of a process-scoped EXIT trap. This avoids any
potential trap scope issues and makes cleanup behavior
explicit and visible at each error path.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Expand the EXIT trap in run_kubernetes_test to also delete the
Kubernetes tenant resource. Previously, early exit paths (exit 1
on node readiness failure, LB failure, etc.) would skip the final
kubectl delete, leaving the resource behind.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
- Add pre-cleanup for VMInstance in second vminstance.bats test
- Fix indentation of LB_ADDR block in run-kubernetes.sh
- Add --ignore-not-found to final Kubernetes resource cleanup
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
- Add EXIT trap in run-kubernetes.sh to clean up port-forward
process and tenantkubeconfig file on any exit path
- Fix LoadBalancer retry loop: success on attempt 20 was falsely
reported as failure due to $i -eq 20 check after break
- Kill stale port-forward in bucket.bats before retry
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
- Add pre-cleanup of stale resources to all app E2E tests so
retries start fresh instead of patching stuck state
- Wait for port-forward to be ready before using it in
kubernetes tests (fixes race condition)
- Reduce version check sleep from 5s to 1s for faster retries
- Clean up temp directory on test failure in cozytest.sh
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Three changes to reduce flakiness in kubernetes E2E tests:
- Increase node Ready timeout from 2m to 5m to accommodate
resource-constrained CI runners
- Fail fast when nodes are not Ready instead of continuing to
LB/NFS tests that will also fail, saving ~7 minutes per attempt
- Delete stale Kubernetes resources at test start to ensure
retries provision from scratch instead of patching stuck state
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
cozytest.sh executes .bats files as plain shell functions where bats
builtins like `run` are not available. Use `!` negation to assert that
readonly user upload fails.
Signed-off-by: IvanHunters <xorokhotnikov@gmail.com>
The mc client requires --insecure on each command when connecting to
SeaweedFS S3 with self-signed certificates via port-forward.
Signed-off-by: IvanHunters <xorokhotnikov@gmail.com>
Update bucket E2E test to match the new per-user access model:
- Create bucket with admin (readwrite) and viewer (readonly) users
- Test that readwrite user can upload, list, and download objects
- Test that readonly user can list and download but cannot upload
- Use per-user BucketAccess and credential secret names
Signed-off-by: IvanHunters <xorokhotnikov@gmail.com>
<!-- Thank you for making a contribution! Here are some tips for you:
- Start the PR title with the [label] of Cozystack component:
- For system components: [platform], [system], [linstor], [cilium],
[kube-ovn], [dashboard], [cluster-api], etc.
- For managed apps: [apps], [tenant], [kubernetes], [postgres],
[virtual-machine] etc.
- For development and maintenance: [tests], [ci], [docs], [maintenance].
- If it's a work in progress, consider creating this PR as a draft.
- Don't hesistate to ask for opinion and review in the community chats,
even if it's still a draft.
- Add the label `backport` if it's a bugfix that needs to be backported
to a previous version.
-->
## What this PR does
### Release note
<!-- Write a release note:
- Explain what has changed internally and for users.
- Start with the same [label] as in the PR title
- Follow the guidelines at
https://github.com/kubernetes/community/blob/master/contributors/guide/release-notes.md.
-->
```release-note
[ci] Added more debug information to ci tests
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Chores**
* Enhanced error handling and diagnostic output in development testing
infrastructure.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## What this PR does
Adds OpenBAO (open-source Vault fork) as a new managed PaaS application
in Cozystack.
**Structure follows existing app patterns (qdrant, nats):**
- System chart with vendored upstream `openbao/openbao` (chart v0.25.3,
appVersion v2.5.0)
- App chart with standalone/HA mode switching based on replicas count
- TLS via cert-manager self-signed certificates per instance
- ApplicationDefinition, PackageSource, PaaS bundle entry
- E2E test with init/unseal workflow
**Key design decisions:**
- `replicas: 1` → standalone mode with file storage; `replicas > 1` → HA
with Raft integrated storage and retry_join with TLS peer verification
- TLS enabled by default — each instance gets a self-signed Certificate
with DNS SANs covering services and pod addresses
- `disable_mlock = true` in HCL config since default security context
drops IPC_LOCK capability
- Injector and CSI provider disabled (cluster-scoped components, not
safe per-tenant)
- No auto-init/unseal — OpenBAO requires manual initialization by design
- E2E test performs full lifecycle: deploy, wait for certificate + API,
init, unseal, verify readiness, cleanup
### Release note
```release-note
[apps] Add OpenBAO as a managed secrets management service with standalone and HA Raft modes, TLS enabled by default
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
## Release Notes
* **New Features**
* Added OpenBAO managed secrets management service with
high-availability and standalone deployment options
* Integrated monitoring and dashboards for operational visibility
* Enabled configurable external access and web UI
* Added automated snapshot backup capability
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
- Kill stale port-forward processes before starting a new one;
on retries, the previous attempt's port-forward still holds the
port, causing all kubectl commands to get "connection refused"
- Use -ge 2 instead of -eq 2 for node count check; MachineHealthCheck
may create a 3rd VM, leading to 3 nodes joining the tenant cluster
which would never satisfy the exact equality check
- Increase node join timeout from 5m to 8m; QEMU VMs with v1.34 need
more time to boot and join when running after kubernetes-latest
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
When running kubernetes-latest and kubernetes-previous E2E tests
simultaneously, worker VMs compete for resources in the sandbox
environment. 3 minutes was insufficient for nodes to boot and
join the tenant cluster under load. Increase to 5 minutes.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Use tenant base domain in default hostname construction (harbor.RELEASE.DOMAIN)
to match the pattern used by other apps (kubernetes, vpn). Remove unused $ingress
variable from harbor.yaml. Add cleanup of stale resources from previous failed
E2E runs.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Replace PVC-based registry storage with S3 via COSI BucketClaim/BucketAccess.
The system chart parses BucketInfo secret and creates a registry-s3 Secret
with REGISTRY_STORAGE_S3_* env vars that override Harbor's ConfigMap values.
- Add bucket-secret.yaml to system chart (BucketInfo parser)
- Remove storageType/size from registry config (S3 is now the only option)
- Use Harbor's existingSecret support for S3 credentials injection
- Add objectstorage-controller to PackageSource dependencies
- Update E2E test with COSI bucket provisioning waits and diagnostics
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Dump HelmRelease status, pods, events, and ExternalArtifact info
when harbor-test-system fails to become ready, to diagnose the
root cause of the persistent timeout.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add postgres-operator and redis-operator to PackageSource dependsOn
to ensure CRDs are available before Harbor system chart deploys.
Make persistentVolumeClaim conditional to avoid empty YAML mapping
when using S3 storage without Trivy.
Increase E2E system HelmRelease timeout from 300s to 600s to account
for CPNG + Redis + Harbor bootstrap time on QEMU.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add registry.storageType parameter (pvc/s3) to let users choose
between PVC storage and S3 via COSI BucketClaim. Default is pvc,
which works without SeaweedFS in the tenant namespace.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
ApplicationDefinition has prefix "harbor-", so CR name "harbor" produces
HelmRelease "harbor-harbor". Use name="test" and release="harbor-test"
to correctly reference all resources.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Move Harbor from packages/extra/ to packages/apps/ as it is a
self-sufficient end-user application, not a singleton tenant module.
Update bundle entry from system to paas accordingly.
Replace registry PVC storage with S3 via COSI BucketClaim/BucketAccess,
provisioned from the namespace's SeaweedFS instance. S3 credentials are
injected into the HelmRelease via valuesFrom with targetPath.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Replace Harbor's internal PostgreSQL with CloudNativePG operator and
internal Redis with redis-operator (RedisFailover), following established
Cozystack patterns from seaweedfs and redis apps.
Additional fixes from code review:
- Fix registry resources nesting level (registry.registry/controller)
- Persist token CA across upgrades to prevent JWT invalidation
- Update values schema and ApplicationDefinition
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add Harbor v2.14.2 as a tenant-level managed service with per-component
resource configuration, ingress with TLS termination, and internal
PostgreSQL/Redis.
Includes:
- extra/harbor wrapper chart with HelmRelease, WorkloadMonitors, Ingress
- system/harbor with vendored upstream chart (helm.goharbor.io v1.18.2)
- harbor-rd ApplicationDefinition for dynamic CRD registration
- PackageSource and system.yaml bundle entry
- E2E test with Secret and Service verification
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Quote all tenantkubeconfig-${test_name} references in run-kubernetes.sh
for consistent shell scripting style. The only exception is line 195
inside a sh -ec "..." double-quoted string where inner quotes would
break the outer quoting.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
When the kubernetes E2E test fails at the deployment wait step, set -eu
causes immediate exit before cleanup. On retry, kubectl apply outputs
"unchanged" for the stuck deployment, making retries 2 and 3 guaranteed
to fail against the same stuck pod.
Add pre-creation cleanup of backend deployment/service and NFS test
resources using --ignore-not-found, so retries start fresh. Also
increase the deployment wait timeout from 90s to 300s to handle CI
resource pressure, aligning with other timeouts in the same function.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Replace `|| echo ""` with `|| true` to avoid newline bugs in variable
assignments. Switch `for x in $(cmd)` loops to `while read` for safer
iteration over kubectl output.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
- Move nil check before req dereference in CreateVolume
- Scope CiliumNetworkPolicy endpointSelector to specific VMI
- Use vmNamespace from NodeId for VMI lookup instead of infraNamespace
- Log PVC lookup errors in ControllerExpandVolume
- Wrap CNP ownerReference updates in retry.RetryOnConflict
- Fix infraClusterLabels validation to check runControllerService flag
- Dereference nodeName pointer in error message
- Replace panic with klog.Fatal for consistent error handling
- Honor CSI readonly flag in NFS NodePublishVolume
- Log mount list errors in isNFSMount
- Reorder Dockerfile ENTRYPOINT after COPY for better layer caching
- Add cleanup on e2e test failure and --wait on pod deletion
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
Remove separate NFS Application dependency from e2e test. The kubevirt
CSI driver wrapper now handles RWX Filesystem volumes natively - PVCs
with ReadWriteMany accessMode use the standard kubevirt StorageClass.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
The mysql chart actually deploys MariaDB via mariadb-operator, but was
incorrectly named "mysql". Rename all references to use the correct
"mariadb" name across the codebase.
Changes:
- Rename packages/apps/mysql -> packages/apps/mariadb
- Rename packages/system/mysql-rd -> packages/system/mariadb-rd
- Rename platform source and bundle references
- Update CRD kind from MySQL to MariaDB
- Update RBAC, e2e tests, backup controller tests
- Keep real MySQL CLI/config tool names unchanged (mysqldump, [mysqld], etc.)
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
## What this PR does
Remove the FerretDB managed application from Cozystack. This includes
the application Helm chart, resource definition, platform source, PaaS
bundle entry, RBAC clusterrole entry, and e2e test. Historical migration
scripts are left intact for upgrade compatibility.
### Release note
```release-note
[ferretdb] Removed FerretDB managed application
```
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Chores**
* Removed FerretDB managed database service and associated Helm chart,
documentation, and test components from the platform.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Keep the 1-minute timeout for other components (cilium, coredns, csi,
vsnap-crd) to preserve fast failure detection, and apply the 5-minute
timeout specifically to ingress-nginx which needs it after the
hostNetwork to NodePort migration.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
