After upgrading vCenter Server and the Supervisor to version 9.0.2, backups that include the Supervisor Control Plane fail. Manual backups that exclude the Supervisor complete successfully.
The /var/log/vmware/wcp/sv_backup_script.log file records errors fetching the auto-attach-service image, similar to the following:
DEBUG backup-service [csp@4413] [backup.py:65] Running ['/usr/lib/vmware-wcp/tools/imgpkg', 'copy', '--bundle', 'localhost:5002/images/vksm/auto-attach-service', '--to-tar', '/var/lib/vmware/wcp/backup/xxxxx.tar']
ERROR backup-service [csp@4413] [backup.py:72] Cmd ['/usr/lib/vmware-wcp/tools/imgpkg', 'copy', '--bundle', 'localhost:5002/images/vksm/auto-attach-service', '--to-tar', '/var/lib/vmware/wcp/backup/xxxxx.tar'] failed. ret=1, stdout=, stderr=imgpkg: Error: Fetching image: GET http://localhost:5002/v2/images/vksm/auto-attach-service/manifests/latest: MANIFEST_UNKNOWN: manifest unknown; map[Tag:latest]
Traceback (most recent call last):
File "/usr/lib/vmware-wcp/backup-restore/backup.py", line 67, in run
The svc-auto-attach-domain-xxx namespace has an incomplete image reference in its vmware-system-service-images annotation. The image URL is missing its tag, which causes the backup script to fail with a MANIFEST_UNKNOWN error when determining which images to back up.
To workaround this issue and allow proper package updates, perform the following steps to reset the auto-attach service:
1. Log in to the VCFA UI and uninstall the auto-attach service from VKSM.
2. Reset addon status in vmsp cluster: this will allow proper update of the package yaml in VCFA
ssh vmware-system-user@$VCF_HOST
sudo su -
kubectl patch addon broadcom.auto-attach-service-0.1.0 -n prelude --type=json \
-p='[{"op": "remove", "path": "/status/serviceUnpackageTask"},{"op": "remove", "path": "/status/downloadTask"}]' \
--subresource=status
3. Reinstall the auto-attach service from the VCFA UI to install it with the correct source tag.
Engineering is aware of it and working on a fix.