Wanted to let everyone know that on Thursday, David Nanian of Shirt Pocket Software sent me this:
“The replicator works in 15.3.”.
My assumption is that he tested SD with the latest beta version, V15.3.2, of Sequoia that was released on Thursday.
So, all is well with bootable clones, at least with SuperDuper!. Sure glad we have some developers around like David who refuse to give up.
Hmmm, I selected Regularly Scheduled Backups because I use CCC; perhaps I could have selected BOTH it and Versioned Backups?
Adam,
I can imagine a large group of users / readers do not understand that you have to back up your cloud data. That would be Google, Microsoft, and Apple.
Today many cannot ensure that all of the data is synced to a local device so that can be backed up. Also systems like Google Workplace and Microsoft 365 do not provide backups of that data…
I think the strategy needs to be more top-down and less “device-centric”
One topic that I’m interested in is in which backup strategies are resilient to having a bad actor hack your system and encrypt all of your data. I don’t know enough technical details to know which (if any) of the suggested backup techniques would be resistant to this.
Wow, you made this way too long and overly complicated. Do you get paid by the word?
All anyone needs to do is Google “3-2-1 backup strategy”. It’s that’s simple. It’s not rocket science.
Also, I skimmed of 90% of this because it was superfluous, but did you fail to mention ChronoSync? I have found that to be the best solution in my 35 years as an expert. It has versioning, and works with cloud storage, too.
Finally, for the hardcore coders, you can get the same functionality just using rsync via the terminal or with scripts. It’s free and built-in to the OS.
A backup that is physically detached after the backup will not be affected by a system hack. If the backup is still attached but not mounted, the hack would need to survey all attached disks (via Disk Utility) or its command-line equivalent and then mount each backup disk (and access it with its encryption key if encrypted) to mess with it.
I use CCC to run scheduled backups. CCC has the encryption key for each encrypted volume. It mounts each volume before running the backup and unmounts it afterward. Setting this up using controls in CCC’s dialogs was easy.
Yes, indeed. This is my default mode of operation. For this reason, I consider “duplicates” basically defunct: you can browse and migrate from TM backups just as you would a duplicate, and duplicates are only bootable when you put a lot of work in to keep them current and Apple haven’t broken a system component they’re clearly not all bothered about. So farewell, duplicates.
Definitely not cloud storage - if malware encrypts your data, then that encrypted data is going to get synced to the cloud.
I would say that a local backup strategy that involves APFS snapshots (like Time Machine or CCC) is a good start. Even if your encrypted data gets auto-backed-up, your should retain older snapshots of everything from before the attack. And snapshots are all read-only, so malware can’t encrypt them (although I suppose it might be able to delete them if it can acquire suitable permissions).
One or more off-line storage mechanisms (e.g. clones that are only connected and powered when you’re making backups) can help protect against malware that deletes backups/snapshots. It can’t delete what isn’t connected.
I’d think the ransomware would have to delete the snapshots, else you’d run out of disk storage while it was encrypting.
Time Machine will auto-delete old snapshots as necessary. And yes, if encrypting (and therefore modifying) everything forces it to delete all of the snapshots, that is going to be a critical problem.
Having at least one off-line (or at least manually-generated) backup is a good protection against this.
Using a backup storage device that is at least twice the size of your data is another good option.
FWIW, my Mac has a 2TB storage device, and I’m using about 1TB of it. My backup devices are 4TB drives. So they can hold 3-4 full backups, and a massive number of incremental backups. I chose this size for my backup media primarily because I want to retain a long history, and also because 4TB HDDs aren’t very expensive, but they also help protect against something that creates massive damage to files.
I agree, which why I explicitly called out that need in the article.
First off, there is no widespread ransomware targeting the Mac, so while it’s slightly beyond a theoretical problem, it’s not a real-world issue.
There is immutable cloud storage—that’s how Retrospect protects against ransomware.
The real question revolves around how long ransomware could remain undetected on the assumption that as long as it’s in control, all data being backed up would be encrypted, rendering those backups worthless.
I think you’re probably right in most cases. But, I just had a terrible case where a Western Digital external HD failed – not the drive itself, but the controller or whatever the circuitry is that allows the drive to communicate. Unbeknownst to me, this drive had built-in encryption, meaning that simply putting the drive in a new enclosure did not make it readable. The decryption was built into the failed circuitry. Western Digital told me they couldn’t help and didn’t have their own recovery service. I was SOL unless I was willing to pay a third party drive recovery firm to do some kind of deep-dive recovery effort that would cost thousands of dollars it if worked at all. I couldn’t justify that for my personal files. (There’s another piece to this involving a lost Backblaze backup of the drive, but it’s not worth going into that horror story here…)
For a lot of users of cloud backup, this is a pretty well-hidden threat to their data. Microsoft OneDrive (which we use at work) defaults to, and pushes pretty hard to encourage, keeping only stubs on the local drive. It and other services like Dropbox try to make this as seamless as possible, which is nice in terms of usability – but it also means that aside from a subtle icon alongside a file, it’s not easy to notice that the file isn’t stored locally. I think(?) that software like Carbon Copy Cloner warns users when it encounters these stubs, but the situation is beyond what a lot of casual users can untangle.
CCC actually has a feature to deal with this situation. There’s an advanced setting called “Temporarily download cloud-only files to make a local backup”.
When enabled, CCC will look for these files. For each one found, it will download the file, make a backup, and then evict the local file.
They recommend that you do not do this as a part of a full-system backup, but instead make a separate backup of your cloud storage, which automatically enables this feature, but the feature exists for those who need it.
There are some gotchas, however, so I recommend reading the linked article (below).
See also: CCC: Backing up the content of cloud storage volumes.
Interesting discussion.
One has to wonder if Apple care enough about the whole backups offering they have, or do they essentially just think that iCloud Drive + Time Machine are going to be their good enough solutions for their customers for the foreseeable future?
While iCloud files are of course duplicated in their data-centres, some form of further in-the-cloud backup service might be a future service to sell, perhaps. Though prices would likely have to come down for their cloud-based storage before it’d be viable, me thinks.
I’ve also considered using Backblaze, but subsequently given up on the idea. While they have good sounding solutions to ‘backing up TB’s of everything you own for one small affordable monthly fee’, problems arise when one has to avail themselves of reliable egress (getting your data from them) when it comes to recovering TB’s of data. Some issues I’ve read a lot about:
Egress (re-downloading) data en-masse online is very difficult and prone to repeated failures, eg. Batches are time-consuming, just 0.5TB each, and prone to failures.
Their 8TB HDDs have max 8 drives/year (i.e. 64TB hard limit), the rest have to be kept thus paid for. Making larger backups difficult and expensive.
Recovery is only available from N.American data-centres. So importing those HDDs incurs import duties/taxes on each one for international users. Getting later refunds on those is then time-consuming.
Back-ups often fail.
Their customer service is difficult to deal with if and when said issues arise.
The Reddit forums (among others) have had quite a bit on this over the years, and so I decided against using them, unfortunately. YMMV.
I had a similar experience with an OWC 4 bay enclosure formatted by SoftRaid. SoftRaid started giving warnings about an imminent HDD failure and so I replaced the HDD. But there were more warnings and more HDD replacements. I swopped HDDs around but still the warnings coming. Then I suspected the HDDs were not at fault and tested them on another non-OWC enclosure and formatted using Apple Utility app - no issues still.
It was the controller in the OWC enclosure that was at fault. OWC was not helpful in either replacing the enclosure nor having an app to test OWC enclosures. SoftRaid was not contrite in acknowledging its software was at fault in giving misleading error messages nor any indications of a software fix.
I have been a personal subscriber to Backblaze services for some time and I am based in Australia and so doing very remote off-site backups. Backblaze has managed to handle my ISP changes, computer and external drives replacements, and modifications to file types, etc. Backing-up is done in the background and I am often surprised how immediate these backups are done. There’s not much latency between the keyboard and the backup.
I have had to some online restores from time to time and I haven’t had the issues as you describe. I did one major restore - nearly 8TBs of image files - and that meant having a drive posted to me from the US. All image files restored worked. It did take a little over a week for delivery from the US to Australia, which was a bit of a surprise. For people outside the US, the US postal and courier services are notoriously slow and expensive (China is the benchmark). The one-year free versioning works, but probably not reasonable to do a big restore.
As far as I know, Backblaze data storage has not been compromised and that’s a key requirement for me.
I have not had to contact customer service, in part because I have not had a reason to do so. But the Backblaze website can be a confusing, difficult-to-navigate mess. But that compliant can be levelled about most online service websites these days.
The price one pays for an online backup service depends on the value you attribute to your files.
One thing to mention, databases are not backed up every time a record is updated, correct? I assume my 72GB FileMaker Pro database of work images is not backed up again each time a new record is added. I periodically do a save the database (Save As (compacted)) to a thumb drive. Smaller databases, thanks to this article, are now saved as compacted onto iCloud so I’m working locally and not on the cloud version. (I should script that!)
I believe that is incorrect…when you make a change FM writes that to the file. The file is in use so might not get backed up to BB until you exit the app though.
Not only did I need to be reminded, I need to be instructed. What you describe is not at all what I see. But first, does it matter that I’m using a Time Capsule? (If it does, and I should check if it’s formatted as APFS, please tell me how to do that. I think I got it in 2015, in case that’s significant.)
In case the Time Capsule issue is irrelevant, I’ll continue. I see a sparsebundle file (actually three, apparently for three different logins, on two computers). For each user, I see something like the following for the package contents (although the single login on the one computer does not have the “mapped” directory).
What am I doing wrong? Thanks for insight.
