7:44 p.m.
Here is how I manually cancel backup operations caught in a finalizing state from Veeam.
I've logged multiple tickets with Veeam about this issue. There is a forum post [1]
about it as well.
1. Obtain a oAuth access token (if keycloak is used).
curl -k -H "Accept: application/json"
'https://<engine>/ovirt-engine/sso/oauth/token?grant_type=password&username=<username>&password=<password>&scope=ovirt-app-api'
Where the <username> and <password> are URL encoded. It can easily be done
with urllib in python.
>> import urllib.parse
>> query = "admin@ovirt@internalsso"
>> urllib.parse.quote(query)
'admin%40ovirt%40internalsso'
>> query = "super!secret"
>> urllib.parse.quote(query)
'super%21secret'
The curl command will return a JSON object, extract the value of the access_token, set it
aside. It's good for a while.
token=eyJhI...FpZpA
More information can be found here:
https://www.ovirt.org/documentation/doc-REST_API_Guide/#authentication
2. Gather the VM IDs of the affected machines from the Engine UI.
vmid=141d2650-a1c6-430b-b364-cc15f83ec50c
3. List the backups for each VM. This will return an XML object. Extract the backup-id
<backup id="...">.
curl --insecure --header "Authorization: Bearer ${token}" --request GET --header
'Version: 4' https://<engine>/ovirt-engine/api/vms/${vmid}/backups
https://www.ovirt.org/documentation/doc-REST_API_Guide/#services-vm_backu...
4. Log into the engine or anywhere the oVirt SDK is installed.
5. Configure the client, check and set the following:
a. Create a file in ~/.config called ovirt.conf.
b. Download a copy of the engine CA to the home directory.
c. Specify the engine URL, username, password, and path to ca.
[engine]
engine_url=https://<fqdn-of-engine>
username=<username>
password=<password>
cafile=ca.pem
The <username> and <password> are not encoded. Ensure the permissions of the
file are set to 0600 or 0400.
6. Use the backup_vm.py script in /usr/share/doc/python3-ovirt-engine-sdk4/examples.
engine ~]$ /usr/share/doc/python3-ovirt-engine-sdk4/examples/backup_vm.py -c
<engine-id> stop <vm-id> <backup-id>
Where <engine-id> matches an entry in ~/.config/ovirt.conf. In this case I called it
"engine".
Once the backup has been cancelled, go ahead and delete any lingering Veeam snapshots.
You should be able to create snapshots as well, but I noticed Veeam backups will
consistently fail if there are any preexisting snapshots. It seems to think there are 0
bytes in the disk image and immediately fail. The only way I found to backup a VM using
the most recent Veeam plugin is to ensure all snapshots are deleted.
The other thing I recently noticed is that ovirt-imageio service eventually stops
responding to requests, resulting in consistent backup failures. Each backup request will
remain in a finalizing state, requiring them to be manually cancelled as outlined above.
After the last round of backup failures, I finally noticed about 14+ connections from the
RHEV Backup Proxy from Veeam (or passed through the engine by Veeam) to each Hypervisior.
The connection state is CLOSE_WAIT. The logs for ovirt-imageio have uncaught exceptions,
connection closed by peer responses. Not sure if this is the reason why they are lingering
around at the application level. However, if there are no backups in progress, go ahead
and restart the service. It might take a few minutes for it to restart. The service should
start responding, all CLOSE_WAIT connections will be gone.
systemctl restart ovirt-imageio
Each time I find something different with the most recent RHEV Plugin from Veeam, so my
advise should be taken with a grain of salt.
Hope this helps!
[1]
https://forums.veeam.com/red-hat-virtualization-f62/veeam-rhv-12-1-comman...