Hm I wonder why snapshots wouldn’t satisfy 3. Copies on the same disk like /file, /backup1/file, /backup2/file should satisfy 3. Why wouldn’t snapshots be equivalent if 3 doesn’t guard against filesystem or hardware failure? Just thinking and curious to see opinion.
Does this make sense?
If Raid is backup, then Unraid is?
Try ZFS send if you have ZFS on the other side. It’s insane. No file IO, just snap and time for the network transfer of the delta.
Every hour. Could do it more frequently if needed.
It depends on how resource intensive the backup process is.
Consider an 800GB Immich instance.
Using Duplicity or rsync takes 1 hour per backup. 99% of the time is spent in traversing the directory structure and checking which files have changed. 1% is spent into transferring the difference to the backup. Any backup system that operates on top of the file system would take this much. In addition, unless you’re using something that can take snapshots of the filesystem, you have to stop Immich during the backup process in order to prevent backing up an invalid app state.
Using ZFS send on the other hand (with syncoid) takes less than 5 seconds to discover the differences and the rest of the time is spent on the data transfer, at 100MB/s in my case. Since ZFS send is based on snapshots, I don’t have to stop the service either.
When I used Duplicity to backup, I would backup once week because the backup process was long and heavy on the disk array. Since I switched to ZFS send, I do it once an hour because there’s almost no visible impact.
I’m now in the process of migrating my laptop to ZFS on root in order to be able to utilize ZFS send for regular full system backups. If successful, eventually I’ll move all my machines to ZFS on root.
What’s the second B stand for?
Unless AI dramatically improves from where LLMs are today (in ways that it so far hasn’t), as a worker, I’m looking forward to the drastic shortage of experienced senior devs in a few years time.
If we look at other companies that have been in Apple’s position, this seems to be a temporary state of affairs even if it lasts for a while. If the expectation is for profit to grow year over year (it is), as growth of market share stalls because you’ve already expanded as much as you could, you’d get pressed to find profits by exploiting existing revenue streams. That’s the point when employee opposition stops working. Think of the recent events when the Google Search VP opposed the Ad VP’s requests to make search worse in order to improve ads revenue. The Ad VP got appointed to lead search and the previous search VP got moved to a dark corner somewhere. Once you run out of profit growth in the existing revenue streams, they’d ask you to find profit growth by reducing labor cost. We also saw that happening in various companies over the last little while.
If Apple was a private corporation owned by some people who aren’t looking for ever increasing profits, I’d believe they might not follow this pattern. But they aren’t.
That’s just my guess and the reasons behind it. Could turn out that you’re right and Apple is an exception to the rule. I mean, I hope it does but I’m not optimistic.
Any replacement is subject to the priorities of Apple’s major shareholders and its board. Those push for further monetization of Apple products and services.
It might be time for the Taiwanese people to take a page from their Chinese comrades and learn that they can’t rely on a private corporation to do what’s in their and their country’s best interest. If they want to keep that democracy, they better reign in their capitalists before they sell the country’s security for profit.
Why did you do that to yourself… 🥹
Interesting. They already make their motherboards in Vietnam.
With 2 disks that would be type mirror in ZFS-speak, completely built-in. Equivalent to RAID1 in terms of hardware fault tolerance.
You could do a 3-disk mirror or n-disk mirror really. The RAID5/6 rough equivalents are called RAIDzN where N is the number of disk failures they tolerate. E g. RAIDz1, RAIDz2, etc. You probably want a mirror unless you need more space than a single disk provides.
Yup, turn it on, let it do a scrub, then turn it off. I’d still use redudnancy though. Not merely to cover the case of the drive failing, but also to cover the bit rot use case. It’s exceedingly unlikely bits to rot at the exact same spot on two or more disks. When ZFS finds a checksum mismatch during a scrub (which indicates bit rot), it’ll be able to trivially recover the data from the drive where the checksum matches. It’ll then rewrite the rotten part.
JFC… This such a trainwreck…
ZFS with automatic snapshots and scrubbing. This will keep as many and as old snapshots as your like. It’ll ensure the files don’t rot. It’ll ensure the media doesn’t die, so long as you have enough redundancy and you replace disks as they die. This is what I’d trust for long term storage because I think I understand how and why it works. It should last as long as I feed it disks. If I delete something, I should be able to restore it from a snapshot. The hardware doesn’t need to be anything fancy. Just a Pi 4/5 with a couple of WD Elements would be fine. Could add more disks for more redudnancy. I’m running 2-disk residency.
You don’t have to touch the software if it’s not exposed to the Internet. Whatever works today on it will work 20 years from now, so long as the hardware works. A couple of spare Pis, SD cards and power supplies should let it last for decades.
Use Debian LTS or Ubuntu LTS (10 years support with free Ubuntu Pro). Turn on automatic unattended updates. Upgrade OS when you’re bored one of those years.
Keywords:
In a statement to the publication, Signal president Meredith Whittaker says, “Our privacy standards are extremely high and not only will we not lower them, we want to keep raising them. Currently, working with Facebook Messenger, iMessage, WhatsApp, or even a Matrix service would mean a deterioration of our data protection standards.”
Ugh, okay Meredith, let’s pretend it’s impossible to handle this with user experience that makes the user acknowledge their conversation with a WhatsApp user is not secure. Meanwhile if the only viable way for this conversion to occur is to have WhatsApp on both ends, the situation less secure. So according to Meredith, the choice is between less overall security or not having conversations with people who don’t use Signal. That could makes sense for her salary but it surely is a net negative for Signal users some of which will have to install WhatsApp since they won’t be able to afford not to have those conversations.
Right so I guess the question of 3 is whether it means 3 backups or 3 copies. If we take it literally - 3 copies, then it does protect from user error only. If 3 backups, it protects against hardware failure too.
E: Seagate calls them copies and explicitly says the implementer can choose how the copies are distributed across the 2 media. The woodchipper scenario would be handled by the 2 media requirement.