Backup question

I'm not aware of any additional technical description, but can probably make an educated guess…

First, whichever host machine you're on has complete access to your Linode, including all disks, so to your first question about access, it's nothing more than the host reading the appropriate device. This is no different from how your disk is initialized for a new distribution or disk image, how cloning works, etc…

Next, the backup system is documented as a filesystem backup that mounts your filesystem (e.g,. as opposed to treating the disk as a block device) and uses snapshots. The document also talks about modification tracking and from personal experience backups tend to require backup times that correlate with the changes since the prior backup (e.g., a typical daily backup when nothing much changes on the node is very fast).

So you can probably model this mentally as a backup process on the host that mounts your filesystem, takes a filesystem level snapshot (think LVM or ZFS) for filesystem consistency - although as the document warns, that may not be the same as application consistency - and then performs an incremental file-based backup between your current filesystem and your most recent backup. Changes are determined strictly by file timestamp and/or size. Under the covers file data is only needed for your currently active set of backup slots and/or snapshot, so I'd assume anything else is discarded.

The only logs you'll have access to are those shown in the first screen shot - basically pass/fail with time information.

From my perspective the backup service is good as a baseline disaster backup, providing you a quick way to restore a system if needed onto a brand new Linode (at least in the same data center) following a catastrophic loss of the active node. Oh, and I think cloning still requires your Linode to be shut down, so a backup snapshot + restore can clone without shutting down. But it's mostly all or nothing - though if you have the space, you could restore, mount the restored disk and be more selective. If you want or need any features above that, you can use an additional backup layer above the Linode backups. For myself, I use the backup service on all of my critical nodes (for fastest recovery in an emergency), and have a separate backup system (bacula) that I run across all systems independently which I use for selective restores and more configurable control over backup lifetimes.

-- David