I'm having trouble understanding if you're doing file folder backups or you're asking about image based backups. You mentioned file backups but then you also talked about synthetic full backups which apply only to image backups. File backups can always be restarted in any way and they'll pick up from where they left off since everything is file based. Image based backups can be restarted but it requires no changes to the volume to properly resume, and that's unlikely on a boot volume. So please clarify and I'll try to update my answer.

I'm using the new backup format, which supposedly does a file-based synthetic (in-cloud copying) backup.

https://help.msp360.com/cloudberry-backup/backup/about-backups/new-backup-format/synthetic-full-backup

You seem to have missed my point, I don't want to have to intervene and restart the backup when interrupted, I want cloudberry backup to resume the interrupted backup itself.

The hypothetical scenario is a machine that is effectively unattended, is only online for a few hours at a time, and a full backup will take 100-ish hours.

At present it appears I have to manually intervene to resume the full backup.

As far as I'm aware backups do not continue automatically if interrupted. They would resume on the next schedule assuming they can be resumed. if not, they would restart the last backup and any unfinished data from the previously failed backup would be removed. I am checking with Support to see if your particular use case is workable with the new backup format. It's possible using the legacy format will be preferable, but I'll let you know when I hear back.

Thanks. I'm a little surprised there is not more demand for unattended restarts given the long term trend towards laptops from always on desktops, and set-n-forget always being cheaper to manage/monitor. I can't imagine looking after a fleet of laptops efficiently without have some kind automatic restart of backups.

if you're concern is that someone might turn off a laptop in the middle of a backup, say at night when they work day ends, and you want to make sure that that backup kicks off immediately when the user starts the laptop the next morning, then you can use the run missed scheduled plan immediately option on the scheduling tab, and also schedule an additional backup to run in the middle of the night, say at 3:00 a.m. . When the laptop boots in the morning it'll detect the 3:00 a.m. backup was missed, and it will continue and start up immediately.

Thanks, David, but I don't think I can make Cloudberry work for for my use case. It's not a work machine, so online hours are not predictable. I did think of your suggestion, but if the machine is left on, it would result in too many full backups and no "down time" between full backups.

It just seems like a little too much monitoring and manual involvement to keep a small number of full backups for this particular machine.

If the extra backup has nothing to back up because the files finished earlier it will detect this and end very quickly without impact to the managed machine. But I’ll enter your request into the system to automatically continue backups after an interruption. Thanks.

Here's a close hypothetical, I'd like one full backup per month. I schedule a 3:00am backup every night to restart after interruptions. I can't see how to get one backup per month without scheduling every night, monitoring for success, and then changing the schedule when I finally get a full backup so as to stop further full backups til next month. Then next month, manually set the schedule to every night again, and so on.... Note: changing the schedule is also a pain as I have to co-ordinate remote to a machine 700 kms away. An option to automatically restart interrupted backups would avoid all this fiddling - just schedule a monthly full backup - done.

Anway, thanks for the help. I'll keep an eye out for changes and re-evaluate again if it looks promising.

If you're performing a file / folder backup using the legacy backup format (which is not going away), the only real full backup is the initial one and that backup can be continued as needed to complete. The "full" as I'm referring to it here is not a a logically linked set of files, but simply a backup of each file. That's why continuing the backup is easy - the product just detects which files have not been backed up yet and backs them up. Subsequent backups, whether they are incremental or full are only backing up changed files. We never back up all files again. Any backups run after the initial full generally run much faster since the breadth of files that are changed is normally low compared to the initial set of files backed up. These subsequent backups also can be continued the next day.

For image backups, however, it's best to have the machine running until completion.

I'm not trying to convince you we are the best solution for you. Just trying to clarify how we work.

If you're using the new backup format, then as I understand it, there may be a bit more work involved to create the synthetic full - compared to the legacy format. I've asked the team for some guidance if my understanding is correct and what customers should expect on duration. But what I would say is that you can use the legacy format and schedule as needed, even multiple times a day since only changed files get backed up, and only the changes within files if you use the block-level option, as that would not be an issue for machines that are not running at all times.

I'll reply back with details on new backup format full backups once I get an answer on those details.

I'm still waiting on some additional information, but I meant to add in the previous reply that in the new backup format, we process files and physically group them together to reduce I/O and improve performance when backing up large numbers of files, especially to cloud storage. We call them "parts". These parts are not large.

In case an upload does not complete because a laptop is shut down, the product will restart (and re-upload the last incomplete part) and then continue with the remaining parts. In that sense the backup effectively continues from where it left off.