For your information, I found the root cause. Our remote backup script would timeout before the save sys ucs command would finish, resulting in temp files in /shared/tmp.
I manually cleaned those, and fixed our script, and everything went back to normal.
One way we deal with this is to trigger the backup worker over the REST API, which is located at the
/mgmt/tm/shared/sys/backup
endpoint.
If you POST to
/mgmt/tm/shared/sys/backup
with a JSON payload of
{'file':,'action':'BACKUP'}
, replacingwith the name of your archive file, you should see a HTTP 202 response if the job was created successfully which will include some details about the job.
You can run a GET query for a running task using the
id
that was returned at the same endpoint:
/mgmt/tm/shared/sys/backup/
and check the status of the job.
Once completed you could retrieve it using SCP if needed.