Hi folks,
(misplaced my earlier login, so starting anew)
Been using various flavours of Linux for ages, primarily server-side, but coming back to the desktop for a primary workhorse after a significant hiatus.
Rolling with Fedora (34) this time, as I find it better suited to BaU work more than 'fun', which is what I need.
Something that I've noticed is that there doesn't seem to be a good backup mechanism baked into the OS, and this is quite the departure from the W10 mechanisms provided through 'File History' & OneDrive. Though I'm not a big fan of pushing sensitive data to a cloud I don't own, manage or control, I've experienced first-hand how this has pulled critical projects form the fire, and the benefits in the way this facilitates something of a seamless experience across desktops is not trivial.
My setups are:
- Desktops - Fedora, Ubuntu, etc; each for different purposes; 90% of endpoints are W10p64 desktops, because that what users/clients/family use
- Servers - baremetal, VM/container hosts & cloud
- IoT/embedded, mobiles
- NAS: FreeNAS/TrueNAS, Synology (paired redundency) & offsite backup; running ZFS & btrfs RAID-5's, making daily snapshots so that I can roll back to an arbitrary day in the last month/year. These hosts also pull by online data via rclone.
- Simple but segmented network with VPN's; no VLAN's or AD/DS
What is really valuable is the way that W10 naively integrates this: users' data backs up to the NAS on the smb://home share, and they can simply right-click on a file/dir of their choice, which correspond to the daily snapshots & "Recover a previous version" with little or no help or intervention on my part.
Because these are server-side generated snapshots, these same snapshots are available to my various POSIX machines over protocol of choice - NFS, iSCSI, SSHFS, etc - but this functionality is not native to the OS AFAICT.
Historically I've managed backups via rsync/clone - typically pulling data (starting with /home & /etc) so that I don't have to install or configure an agent on every client/endpoint - but this is inelegant, and does not effectively use the resources at disposal and introduces complexity that could otherwise be avoided.
This is my ask: what is the more elegant & efficient backup solution?
But I have a few considerations I'm taking into account:
- needs to be Open Source & (near-)native: in the stable repo's and/or in Snap/Flatpak stores.
- Support for FS-based snapshots (btrfs, ZFS, LVM) as well as backing/syncing up that (incremental) data to the network. (if/when my disk dies, my backups die with it unless archived elsewhere)
- needs to be active/current - i.e. git repo still a project under active development/maintenance/activity in the last ~year, not stale
- needs a usable/functional GUI - desktop but web or TUI/ng preferable. I've set this as an arbitrary bar for where I invest time & effort, as I've found this to be a fair indicator of a user-interacting app stack; especially if/when I don't want to do everything for everyone.
Some of of the solutions I've found may be OK fits, but have some shortcomings:
- rclone(baseline reference)
- preferred over rsnapshot & rdiff-backup
- Duplicati (solid choice, but no FS snapshot)
- Snapper (fair candidate)
- Back In Time (stale?)
- TimeShift (doesn't seem to support networking)
- DejaDup(not something I would consider 'elegant')
- based on Duplicity (no FS snapshotting)
- Bacula ('open core', which just sits weird with me...)
- Kopia (intriguing...)
- BackupPC (no FS snapshotting AFAICT)
I know this is a tall order, but if I'm going to invest time & effort into adopting a strategy or solution more widely, I'd prefer it to meet my needs, so that ongoing deployment & use is somewhat uniform, regular & predictable and not needing to contend with numerous exceptions & edge-cases.
I'd appreciate folks insights & comments re this, please.
