* Wait for CRDs to be available and ready
When restoring CRDs, we should wait for the definition to be ready and
available before moving on to restoring specific CRs.
While the CRDs are often ready by the time we get to restoring a CR,
there is a race condition where the CRD isn't ready.
This change waits on each CRD at restore time.
Signed-off-by: Nolan Brubaker <brubakern@vmware.com>
* Pruning unknown fields
In CRD apiversion v1beta1, default preserveUnknownFields=true.
In CRD apiversion v1, the preserveUnknownFields can only be false.
Otherwise, the k8s validation bumps out error message for the
invalid preserveUnknownFields value.
Deploy Velero on k8s 1.16+ with CRD apiversion v1beta1, the
k8s cluster converts apiversion from v1beta1 to v1 automatically.
Fully backup and restore the cluster, restore bumps out error message
due to the preserveUnknownFields=true is not allowed on k8s 1.16+.
Since the CRD structural schema had been defined, enable the preserveUnknownFields
to false to solves the restore bumps out error message on k8s 1.16+.
Signed-off-by: JenTing Hsiao <jenting.hsiao@suse.com>
* Add changelog
Signed-off-by: JenTing Hsiao <jenting.hsiao@suse.com>
Velero client config file should have restricted file permissions to be
read/write-able for the user that creates it--similiar to files like
`.ssh/id_rsa`
Refer to OTG-CONFIG-009: Test File Permission
> Impoper file permission configuration may result in privilledge
escalation, information explousure, DLL injection, or unauthorized file
access.
Therefore, files permission must be properly configured with minium
access permission by default.
[source](https://www.owasp.org/index.php/Test_File_Permission_(OTG-CONFIG-009))
Ticket: #1758
Signed-off-by: John Naulty <johnnaulty@bitgo.com>
* Check for nil LastMaintenanceTime in dueForMaintenance
ResticRepository.dueForMaintenance causes a panic in the velero pod
("invalid memory address or nil pointer dereference") if
repository.Status.LastMaintenanceTime is nil. This fix returns 'true'
if it's nil, so the repository is due for maintenance if LastMaintenanceTime
is nil *or* the time elapsed since the last maintenance is greater than
repository.Spec.MaintenanceFrequency.Duration
Signed-off-by: Scott Seago <sseago@redhat.com>
* changelog for PR#2200
Signed-off-by: Scott Seago <sseago@redhat.com>
* update revision of go-hclog to match go.mod requirement
Signed-off-by: Steve Kriss <krisss@vmware.com>
* update prometheus dep to prepare for go module migration
Signed-off-by: Steve Kriss <krisss@vmware.com>
Install restic with CPU/Memory limits is optional.
If velero cannot parse resource requirements, use default value instead.
After that, the administrator won't get confused that something recovered failed.
Signed-off-by: JenTing Hsiao <jenting.hsiao@suse.com>
Migrate logic from NewUUID function into the pvRenamer function.
PR #2133 switched to a new NewUUID function that returns an error, but
the invocation of that function needs to happen within the pvRenamer
closure. Because the new function returns an error, the pvRenamer should
return the error, the signature needs to be changed and the return
checked.
Signed-off-by: John Naulty <johnnaulty@bitgo.com>
satori/go.uuid has a known issue with random uuid generation.
gofrs/uuid is still maintained and has fixed the random uuid generation
issue present in satori/go.uuid
Signed-off-by: John Naulty <johnnaulty@bitgo.com>
* restic: don't try to restore PVBs with no snapshotID
Signed-off-by: Steve Kriss <krisss@vmware.com>
* changelog
Signed-off-by: Steve Kriss <krisss@vmware.com>
* restic: use restore cmd's env when getting snapshot size
Signed-off-by: Steve Kriss <krisss@vmware.com>
* restic: remove code that considers 0-byte backups an error
Signed-off-by: Steve Kriss <krisss@vmware.com>
* add --allow-partially-failed flag to velero restore create
Signed-off-by: Steve Kriss <krisss@vmware.com>
* remove extraneous client creation
Signed-off-by: Steve Kriss <krisss@vmware.com>
* add godoc to helper func
Signed-off-by: Steve Kriss <krisss@vmware.com>
* todo
Signed-off-by: Steve Kriss <krisss@vmware.com>
Related issue: https://github.com/heptio/velero/issues/1830
This accomplishes everything
that's needed, although there might be room for improvement in avoiding
a GET call for matching CRDs for each resource backed up. An alternative
could be a single call to get all CRDs prior to iterating over resources
and passing this into the backupResource function.
Signed-off-by: Scott Seago <sseago@redhat.com>
* feat: add azure china support
Signed-off-by: andyzhangx <xiazhang@microsoft.com>
* remove AZURE_CLOUD_NAME from required env var fetching
Signed-off-by: Steve Kriss <krisss@vmware.com>
* minor simplification of parseAzureEnvironment
Signed-off-by: Steve Kriss <krisss@vmware.com>
* changelog
Signed-off-by: Steve Kriss <krisss@vmware.com>
* remove cloudNameEnvVar from getRequiredValues call
Signed-off-by: Steve Kriss <krisss@vmware.com>
* just check for err != nil
Signed-off-by: Steve Kriss <krisss@vmware.com>
* Allow the velero server to be created on GCP even without a provided service account key in order to support workload identity and default compute engine credentials. Add option for adding service account annotations.
Signed-off-by: Joshua Wong <joshua99wong@gmail.com>
