harvester / harvester Goto Github PK

Description:
In the UI the image status showing as active when uploading image was failed, e.g

status:
  appliedUrl: https://launchpad.net/cirros/trunk/0.3.2/+download/cirros-0.3.2-source.tar.gz
  conditions:
  - lastUpdateTime: "2020-08-18T10:43:46+08:00"
    message: started image importing
    reason: 'Get http://example.com:9000/vm-images/?location=: dial tcp 93.184.216.34:9000:
      i/o timeout'
    status: "False"
    type: imported
  downloadUrl: ""
  progress: 0

Expected:
change status to failed and display the error message.

Description:

user should be able to config the number of VM instances they want to create on the creating page

Description:
no bootable disk found after creating VM using CDI volume.

When the default version is set for the template, the version will be cleared

add USB-tablet enhancement to the QEMU config that allowing the mouse to report its position without having to grab the mouse.
https://www.qemu.org/docs/master/system/usb.html

This is a feature request:

Add a CRD to store user-customized ssh-keys, the user will need to specify the key name and public SSH key.

add a serial console to the VM using /apis/subresources.kubevirt.io/v1alpha3/virtualmachineinstances/console

Improvement:

The meaning of the env names of VM storage service is not clear enough, it should be compatible with other S3(Object storage) services.

Reference:

Clarify the env names of VM storage service, the following is a feasible proposal.

• MINIO_ACCESS_KEY -> VM_STORAGE_ACCESS_KEY
• MINIO_SECRET_KEY -> VM_STORAGE_SECRET_KEY
• MINIO_URL -> VM_STORAGE_ENDPOINT

TODO:

• Adjust Harvester.
• Change the envs of installation charts.

when trying to run ./bin/harvester, --image-storage flags are not set, and there is little explanation of what they should be set to.

what should they be set to in order to run harvester?

Problem:

When a user deletes an image from the UI the API shows the object is successfully removed, but the WebSocket will continue to push the deleted object to the frontend.

Expected Solution:

WebSocket should not continue to pushing deleted objects.

TODO
remove go vendor but still add go mod tidy

add image metadata ns:name to the dataVolume template annotations when creating the VirtualMachine, therefore user/UI will be able to know which image the URL is related.

apiVersion: kubevirt.io/v1alpha3
kind: VirtualMachine
metadata:
spec:
  dataVolumeTemplates:
  - metadata:
      annotations:
        harvester.cattle.io/imageId: default:myimage ## add image metadata annotation with `ns:name`
      name: lawr-ubuntu
    spec:
      pvc: {...}
      source:
        http:
          url: http://minio.default:9000/vm-images/xxxxx
  template:
    metadata:
      creationTimestamp: null
      labels:
        kubevirt.io/vm: lawr-ubuntu
    spec: {...}

when add a new vm, state is not refresh

set QEMU boot-order (virtual hard drive first and CD-ROM second)

Description:
add embedded/system templates that do not require pre-existing images

• iso-img-template
• windows-template
• qcow2/img/raw-template

need to add Drone CI for the harvester-ui

Issue:
currently, the display name is auto-generated by its URL name and we don't validate whether its name is duplicated or not, the backend should add validation to this.

Solution:
The backed API should validate the display name and auto-increment the display name by 1. e.g os-aaacc to os-aaacc1

few harvester-UI designs and feature changes:

1. vm delete logo
2. vm adjust the location of the namespace
3. networking todo
4. rootdisk can get source: image ，size： imageSize
5. vm add ssh key default select
6. start vm default true
7. template delete namespace (set default namespace)
8. ssh key add a search input
9. ssh key name set the default name
10. vm template option is hidden by default (rancher rke)
11. ssh key The content disappears after input
12. image action launch
13. add disk spell error
14. template describe position adjustment
15. select all position
16. storageclass need to get the first default through API, it cannot be written to hostpath

This is a feature request:

• add image CRD to store VM images by URL
• support Minio as backend server
• add action handler that allows user to upload an image directly from the local computer[low priority]
• add image secret reference

Description:
improve image file validation, check whether CDI or QEMU have tools to validate the correct image type and content

add unit testing cases to the VM template and template version

Steps to replicate:

1. Select one of the images, in this case, select CentOS-7-x86_64-GenericCloud.qcow2.xz
2. Fill up other information and save it

Result:
The virtual machine cannot be created because of the invalid hostname

Suggestion:
It makes no sense that a user cannot create a virtual machine by using a generated name by default even though a hostname is not allowed to have any upper case.

If there is no way to have this kind of name to be saved, it's better to let the frontend convert the uppercase to lowercase before sending it to the backend.

This is a feature request:
Description:
user should be able to add a VM template with customized virtual machine spec, also it should be able to add a image that belongs to this template.

• add VM template that stores kubevirt virtualMachine spec
• VM template will contain a image reference (i.e image namespace:name)

Improvement:

An entry verification for VM storage service endpoint blocks to start the Harvester, it should delay the verification or support to customize it in bootstrapping.

Reference:

The error report from verification:

I not find the vm ui,this repository is a plugin for rancher?

Be able to show the format when a VM image is imported.
The incentive is to handle root volume differently for iso image.

Improvement:

It's necessary to add testings to control the quality of Harvester.

Reference:

• Construct the unit test cases based on the fake generated clientset. (TDD style)
• Construct the integration test cases based on the Ginkgo. (BDD style)

TODO:

• Adjust the test script of dapper to support two kinds of test.
• Add unit test cases.
• Add integration test cases.

Description:

Add action handlers to the kubevirt virtual machine object.

• start VM action handler
• stop VM action handler
• restart VM action handler
• pause VM action handler
• unpause VM action handler
  - [ ] terminate the VM action handler(e.g permanent delete the VM after 2 minutes?) [TBD]

VM is pending with booting from hard disk when using minimal-cloud image (e.g. https://cloud-images.ubuntu.com/minimal/releases/focal/release-20200729/ubuntu-20.04-minimal-cloudimg-amd64.img)

add unit testing cases to the image and key-pair

Description:

allow the user to manage VM through subresource APIs - e.g pause, unpause, reboot, rename, guestosinfo

• allow VM to reboot through API action handler.
• allow VM to pause/unpause through API action hander.

add vm e2e and integration testing framework

Issue:
the ssh-key controller should auto-generate the fingerprint when uploading a new ssh key.

Expected solution
the fingerprint should be added automatically either user created from the UI, API or kubectl.

Description:

1. use an object-oriented CRD name instead of a generic name.
2. rename API to use harvester.cattle.io instead of vm.cattle.io.

add rancher VM UI with the following scope

• add image management page that allows user to view, upload and edit OS images
• add ssh-key management page that allows user to view, create and edit ssh keys
• add VM listing/view detail page
• add a creating page of VM
• add PV/PVC management page
• add VM template page
• add Graphical Console (VNC)

It would be nice to show the root volume size when a VM image is selected.

add rancherVM helm chart, the current chart should include:

• rancher harvester
• Minio server
• kubevrit
• kubevirt CDI

Issue:
kubevirt's datavoulems and cdiconfigs keep updating its managedFields.time field and it leads to steve keeps refreshing its loaded schema.

Reproduce steps:
kubectl get crds -w and the following two CRDs keeps updating

datavolumes.cdi.kubevirt.io                                     2020-07-15T05:52:53Z
cdiconfigs.cdi.kubevirt.io                                      2020-07-15T05:52:53Z

check the VM server logs and it will print refreshing schema message in every 2-3 seconds:

INFO[0013] Refreshing all schemas                       
INFO[0014] Refreshing all schemas                       
INFO[0014] APIVersion /v1 Kind Bindin
...

Expected:
there are two ways to address this issue:

1. since the managedFields.time is introduced in k8s 1.18 and only used as object metadata, it should not be updated continuously - preferred.
2. ensure steve only refreshing updated CRD schema instead of all.

Import an image from https://launchpad.net/cirros/trunk/0.3.0/+download/cirros-0.3.0-x86_64-disk.img

Result:

ERRO[4747] error importing image from https://launchpad.net/cirros/trunk/0.3.0/+download/cirros-0.3.0-x86_64-disk.img: strconv.Atoi: parsing "": invalid syntax

The server hosting the image file may return a chunked response withoutContent-Length header. In this case, the total file size is unknown until the import is finished. The percentage progress will be unavailable to the users. We might need to show the downloaded size instead(similar to chrome's download behavior).

Description:
When the user chooses to boot MV from the .iso image by default the CD-ROM will be set as the 1st bootable disk in its VM config if the root disk is empty, and after the OS is installed successfully we should allow unmounting the CR-ROM disk easily through the action handler.

manual process:

1. confirm the OS is installed successfully to the VM.
2. kubectl edit vm and remove the cdrom disk and volume.
3. delete the VMI to reboot it.
4. check the VM is running and only the root-disk is enabled.

example of YAML:

1. before remove

apiVersion: kubevirt.io/v1alpha3
kind: VirtualMachine
metadata:
  name: gc-ubuntu-server
  namespace: default
spec:
  dataVolumeTemplates:
  - metadata:
      name: gc-simple-iso
    spec:
      pvc:
        accessModes:
        - ReadWriteOnce
        resources:
          requests:
            storage: 10Gi
      source:
        http:
          url: http://172.16.0.254:8080/ubuntu-18.04.4-live-server-amd64.iso
  - metadata:
      name: gc-simple-root
    spec:
      pvc:
        accessModes:
        - ReadWriteOnce
        resources:
          requests:
            storage: 10Gi
      source:
        blank: {}
  running: true
  template:
    spec:
      domain:
        cpu:
          cores: 1
        devices:
          disks:
          - cdrom:
              bus: sata
              readonly: true
            name: cdrom-disk
            bootOrder: 2
          - disk:
              bus: virtio
            name: datavolumedisk1
            bootOrder: 1
        resources:
          requests:
            memory: 2048M
      volumes:
      - dataVolume:
          name: gc-simple-iso
        name: cdrom-disk
      - dataVolume:
          name: gc-simple-root
        name: datavolumedisk1

2. removed iso disk

apiVersio: kubevirt.io/v1alpha3
kind: VirtualMachine
metadata:
  name: gc-ubuntu-server
  namespace: default
spec:
  dataVolumeTemplates:
  - metadata:
      name: gc-simple-root
    spec:
      pvc:
        accessModes:
        - ReadWriteOnce
        resources:
          requests:
            storage: 10Gi
      source:
        blank: {}
  running: true
  template:
    spec:
      domain:
        cpu:
          cores: 1
        devices:
          disks:
          - disk:
              bus: virtio
            name: datavolumedisk1
        machine:
          type: q35
        resources:
          requests:
            memory: 2048M
      volumes:
      - dataVolume:
          name: gc-simple-root
        name: datavolumedisk1

Description:
While uploading a new ssh-key, the UI should auto-generate the ssh-key name by its public key

Description:

1. user should be able to install harvester chart from the rancher catalog(helm v3)
2. user should be able to access harvester UI through rancher magic proxy
3. allows users to enabled/disable each component i.e Kubevirt, CDI and Minio
4. after deleting the harvester chart from the catalog apps, it should clean all its component like CRDs and CRs
5. when Minio is disabled, s3 server endpoint and secret configuration should be required （need to validate if it works with s3 compatible storage server e.g, AWS s3 or GCS）
6. persistence storage should be enabled by default of the Minio

Description:
both schema validation and generation are missing for the harvester CRD, better to add it since it will add properly schema check.

usage please refer to https://book.kubebuilder.io/reference/markers.html

• 1. Create a VM with no image selected, the create button turns to created - Validated
  

• 2. Create a VM with the only image selected, the page title disappears - Validated
  

• 3. An image just can be searched by using the accurate criteria, e.g., if a user searches the image by using the keyword cent , it will just filter out all the items contains cent accurately. (Case sensitive) - Validated

• 4. Choose a Size, CPU and Memory textboxes are allowed to fill in non-numeric values if a user is using a Chinese input method - Validated

• 5. Choose a Size, CPU and Memory textboxes are allowed to fill in a super large number, and they finally turn into infinity - Validated
  

• 6. Add a disk, there is no limit length on the Name, users can name it with a very very long string. Size has no limit as well, after clicking the show advanced, there is show advanced still displaying. Edit a disk, it still pops up a 'Add a disk' prompt - Validated
  

• 7. Add a networking interface, there is no limit length on the Name and Networking Name, users can name it with a very very long string, size has no limit as well. Edit a networking interface, it still pops up a 'Add a networking interface' prompt. An invalid Mac Address still can be saved - - Validated
  

• 8. Add an SSH key, the length limit for the Name is better to be set in the front end. the SSH-Key label overlaps the content in the textbox. - Will keep tracking it
  

Description:
allow the user to persist data volumes while deleting the VM so it can be reused later.

TODO:

1. need to add harvester UI integration with loaded from the CDN.
2. add embedded UI to the harvester docker image, the user can config to use embedded UI through the ui-index setting with value: local

Description:
set default template version when a new template is created