My wife’s computer is a unibody, late-2012 Mac Mini, model A1347 with Fusion Drive, which we’ve had since then, having ordered it to our specs directly from Apple. We’re happy with it. It’s a lovely little computer with more than enough oomph for my wife’s needs (she is an author).
The HDD on her Mac was silently failing and her computer was getting slower. A quick disk speed test revealed that its write speeds had decreased by about 75%.
Running First Aid on the system volume did not yield any insights into the HDD’s true state. Thankfully, there’s a little app called DriveDx, which I talked about in a previous post. Running that app revealed the HDD’s problems.
The SSD wasn’t doing too well either, but at least its lifespan was at about 50%.
The solution was simple: I needed to replace the HDD. A 1TB SSD would suffice, so I ordered one (an ADATA SU800 1TB SSD). My wife continued to use her computer as usual, since it was still working, although I made doubly sure that it was backing up to Time Machine. I would restore her data from those backups after I replaced the HDD.
Once the SSD arrived, I got to work. I didn’t want Ligia to experience an outage longer than a few hours, so the pressure was on. My plan was to open up her machine, clean the insides thoroughly of dust, replace the thermal paste on the CPU and GPU, then replace the HDD with the SSD. After putting it back together and booting up, I would need to do a data restore.
Here is a gallery of photographs from that process. The insides were indeed full of dust and the thermal paste had dried up. I followed this guide from iFixit, although I have to say it’s not entirely accurate, as detailed below.
I was on my own when it came time to work on the AirPort/Bluetooth board, where the setup differed quite a bit from the guide. There were also a few screws whose location was different in the guide. So I took photos before I disassembled things, just to be safe.
While I love the design of the Mac Mini (inside and out) and I think it’s a fantastic little computer, it’s tricky to work on. Everything has to fit together just right. The things that gave me problems when it came time to re-assemble it were:
the minified SATA cables, which kept popping out of their slots on the motherboard and are really only held in place by the cowling (the little piece of plastic in a semilune shape),
re-seating the top drive, whose side screws have to slide into some holes in the back of the case, but there is little to no tactile feedback when they’re in place, and there’s no way to check things visually; it actually fits asymmetrically over the bottom drive, which is a bit illogical, but that’s how the engineers worked out the hardware design,
and the antenna plate. Oh wow, the antenna plate was a chore to work back in… It has to fit in just right, hugging the inside edge of the case with an indentation made in the wire mesh from which it’s constructed, and for some reason, it just didn’t want to go back in properly. It was off by less than 1 mm, yet it meant that I couldn’t put the screws back on. Be careful with that one!
When it was time to boot it up, the Mac Mini refused to do it. I stared at a black screen for a minute or two, wondering if I’d forgotten to connect some cable inside it, and then it occurred to me to re-seat the AC cable, which is notoriously hard to plug and unplug on this machine, because its slot is too tight. That turned out to be the problem. Whew.
Another wrinkle that I ran into was the Fusion Drive. This machine has an actual SSD inside of it, not a blade SSD, which is what you might find in an iMac or a MacBook. That was a bit of a surprise to me. Anyway, come time to reformat the drives, I figured I could re-enable Fusion Drive and end up with a single volume that used both the Apple SSD and the new ADATA SSD. Nope. While you can run the commands in Terminal to “marry” the two SSDs into a Fusion Drive (see this post for the details), checking the resulting volume with Disk Utility gives an error and Mojave refuses to install on it. So… no Fusion Drive for my wife, I guess. Then I figured I could create a software JBOD in Disk Utility to end up with a single volume once more, and I did that, and it worked, but once again, Mojave refused to install on it. So I had to simply format each SSD as a separate drive and use the 1TB SSD as the system volume, leaving the 128GB Apple SSD as a secondary volume to be used occasionally.
A quick check with DriveDx showed me that the new SSD was doing just fine.
And a disk speed test showed things were humming along nicely.
Here are some Geekbench scores for good measure.
My wife’s pretty happy with it now, she says it is faster than before and it doesn’t crash anymore, which it used to do every now and then. And if my wife’s happy, then I’m happy.
In a recent post, I wrote about upgrading the original (failing) blade SSD in my iMac to a bigger and faster NVMe module. During that upgrade process, I wondered whether splitting my Mac’s Fusion Drive would result in better performance, but decided against it for simplicity’s sake.
Even though I decided against splitting my Fusion Drive at that time, I read articles that advocated for it and suggested even better performance was to be had by allowing the SSD and HDD to run as separate volumes. The idea is to install the OS and select files and folders on the SSD, with the bulk of the files on the HDD. For the sake of experimentation and learning something new, I decided to tinker with my iMac and see if I could squeeze out some extra speed.
For those who are wondering what I’m talking about, Fusion Drive is an Apple technology built into macOS that creates what is essentially a hybrid drive, by combining an SSD module (NAND flash) with a traditional HDD (platter drive) and presenting the two as a single volume to the user. The protocols that govern the data I/O are called Core Storage. Apple writes: “Presented as a single volume on your Mac, Fusion Drive automatically and dynamically moves frequently used files to flash storage for quicker access, while infrequently used items move to the high-capacity hard disk. As a result, you enjoy shorter startup times and — as the system learns how you work — faster application launches and quicker file access.”
I’ve been using Fusion Drive since it came out, retrofitting my iMac at the time with a new blade SSD and thus making it run faster than its original specs. I love this technology, because it offers significant performance improvements for a fraction of the cost of buying a large SSD, which used to be be quite expensive a few years ago.
The long and the short of it is that it’s not worth it to split your Mac’s Fusion Drive. If you’re currently running Fusion Drive on your Mac, keep doing that, you won’t see any significant performance improvements if you split it. Actually, some things may run slower than before, and you’ll also have to deal with a few inconveniences, as detailed below.
I’ll present both scenarios here and you can decide what to do for yourself. There are multiple methods to it. These are the methods I’ve chosen. The number of Terminal commands that you have to run for either scenario is minimal, and the time involved has to do mostly with backing up your computer, waiting for the OS to reinstall and for your data to be restored from backup. For example, if you’ve got a 3TB drive and you’re at about 50-60% usage (and you should be at that threshold or lower on any hard drive), then you should figure on 4-5 hours for either of the two scenarios.
How to split your Fusion Drive
First and foremost, did you backup your computer? If you did, go ahead and create a bootable drive using Apple’s instructions, then boot into it by pressing the Option key as soon as your Mac restarts and holding it down until you see the Apple logo. You need to boot into a separate drive because you’ll be deleting your internal drives entirely, including the boot and recovery partitions.
Once you’re in, open Terminal and get a listing of your disks and volumes.
diskutil list
Your Fusion Drive presents itself as a logical volume group that appears as a separate disk with an HFS+ or APFS partition. Say your SSD is disk0 and your HDD is disk1, your Fusion Drive would be disk2 or disk3. In my case, it was disk3 (disk2 being the bootable recovery drive). Now unmount your internal disks.
diskutil unmountDisk disk0
diskutil unmountDisk disk1
You’ll want to delete that entire disk containing Fusion Drive. Be forewarned, this deletes all you data. Did you backup your computer?
diskutil apfs deleteContainer disk3
Now that Fusion Drive has been nuked, you’ll still have your separate drives that you’ll want to make sure are erased. The eraseDisk command requires that you offer a new name for each disk, so I chose to name them SSD and HDD, to keep things simple.
diskutil eraseDisk JHFS+ SSD disk0
diskutil eraseDisk JHFS+ HDD disk1
Now you’ll want to do a fresh install of macOS onto the SSD, and after that’s complete, you’ll boot up into your fresh install and go to Utilities/Migration Assistant, in order to do a selective data restore. Here you’ll have to decide for yourself, based on the total size of your SSD and your data set, how much of it you’ll want to restore onto the SSD. The rest you’ll need to copy manually from the backup drive onto the HDD. In my case, I restored my user settings and the system and libraries folders onto the SSD, and I copied the following folders onto the HDD: Documents, Downloads, Movies, Music, Parallels (in case you’re running some kind of VM software) and Pictures. Each of those folders was too big to keep on the SSD, even though I have a 512GB module (remember the rule about keeping your drive at or below 50-60% usage).
Once you complete all that work, you’ll need to create links to these folders on the HDD in place of your folders on the SSD. Mojave won’t let you do this when you’re logged into your account, so you’ll need to boot up into recovery mode and open Terminal once more.
Go to your homefolder on the SSD.
cd /Users/yourusername
Delete the folders that are now present on the HDD. You’ll need to do this for each folder that you’ve moved there. Hopefully you’ve written down their names ahead of time.
sudo rm -rf foldername
In your homefolder on the SSD (same location as above), make links to the folders on the HDD. I chose to put mine at the drive’s root level. You may choose to put them in a folder. Just don’t give it the same name as your username, I hear that may cause problems. You’ll need to do this for each folder.
ln -s /Volumes/HDD/foldername
That’s it, restart and use your computer. However, you may find a few inconveniences — these are the ones I experienced:
I noticed no performance improvements. There wasn’t even an improvement in the bootup time. Nothing, nada, zilch.
While Apps may open up faster, if they’re still accessing files on the HDD, editing will still be sluggish. In order for you to see that performance boost talked about with SSDs, both apps and their files need to be on the SSD.
In my case, I had to keep the Photos library on the HDD, because it was too big to keep on the SSD, and while Photos may have opened up fast, loading up the library took forever, until enough of the recent photos were cached on the SSD to allow me to work with my library. So things were a LOT slower with this app.
I kept my mailboxes on the SSD so I was hoping for better performance from Mail, but I didn’t get it. I have a lot of mail stored locally, so in theory, things should have worked faster because everything was on the SSD, but they didn’t. I also experienced odd issues, like when moving messages between mailboxes, it took a lot longer and sometimes didn’t register. I’d drag and drop them, then come back to the app a little while later and find them in the same place, just as if I hadn’t moved them.
iCloud would display an odd notification icon, but when I’d go into it, there was no message. This icon was displayed continually for as long as my Fusion Drive was split. See the screenshot below.
While Time Machine will backup both internal drives, data restores will only restore the files from the SSD. I don’t know why and I don’t know how to fix that, so keep this limitation in mind. You can go into the Time Machine drive manually and copy the files over afterward, but if you run a restore operation on your computer and you wonder where most of your stuff is after it’s completed, don’t freak out, just know you’ll need to get it manually from the drive.
See that “1” over iCloud? It was there all the time.This is the kind of performance the SSD provided when my Fusion Drive was split. It looks impressive, but hold on until you see the same test with Fusion Drive enabled, later down in this post.
How to enable your Fusion Drive
After about a week of running my Mac with a split Fusion Drive, I’d had enough and decided to re-enable it. Here’s how I did it. Before you proceed with this, I’ll ask you again, did you do a full backup of your computer? This will wipe all your data.
Using the same bootable drive, I booted into it and opened up Terminal. Since you’ll be wiping all your internal drives again, you need to be booted from an external drive.
Apple recommends this single Terminal command that is supposed to do everything in one fell swoop. It didn’t work for me, perhaps because my SSD module was a newer NVMe running off an adapter card, not the Apple-approved blade SSD manufactured specifically for this kind of thing.
diskutil resetFusion
I had to do it with a few more commands. First, find out your disk IDs.
diskutil list
Now unmount your internal disks.
diskutil unmountDisk disk0
diskutil unmountDisk disk1
Then create a merged virtual hard drive with Core Storage.
diskutil coreStorage create Macintosh\ HD disk0 disk1
Now get its logical volume group name (the very long alphanumerical name that appears in Terminal after you type this command).
diskutil coreStorage list
Now format and create the JHFS+ volume that will run Fusion Drive.
diskutil coreStorage createVolume yourlogicalvolumegroupname jhfs+ Macintosh\ HD 100%
Don’t worry about formatting the drive to APFS. That’ll happen automatically when you install Mojave. Besides, APFS is not an entirely separate file system, it’s a container running inside HFS+, so like I said, don’t worry about it. That’s it. Now quit Terminal and do a full restore from Time Machine, but prepare yourself for an incomplete data restore (see the reasons given in the previous section). Once the data restore is complete, you’ll need to manually copy the folders that are missing from the Time Machine drive. Or, as I did, you can do a full restore to a backup set that existed before I split my Fusion Drive, which means you’ll get all your old data back in all the right places, but you’ll still need to get your newer files manually from the Time Machine drive.
In my case, I needed to copy the mailboxes, which are located in ~/Library/Mail/V6 from the newest backup set (the one with the split drive) to my computer, and that gave me all my mail, including the interim stuff. I also copied the latest Photos library, and that gave me all my photos, including the interim stuff. Then I went through the Documents and Downloads folders on the Time Machine drive, sorted by date modified and copied the interim files onto my computer. I didn’t need to go through the other folders because I knew I hadn’t worked on other stuff. And once I did this, my data restore was complete. Mail and Photos still needed to rebuild their libraries though, and that took a while.
And because I use Backblaze to backup my computer offsite, I also needed to uninstall and reinstall that, then inherit a previous backup state (don’t worry about this if you’re not using Backblaze).
When that was done, Backblaze told me it had “made” my computer inherit my backup state, as if it had forced it to do this, in a non-consensual way. Kind of a funny way to word things, but their service works well.
Here’s the kicker. I ran another drive performance test after all this, and these were the results.
Actually a little bit faster than before 🧐
Everything runs fast now, and it runs as expected, without hiccups.
As I said at the start of this article, if you’re already running Fusion Drive, do yourself a favor and leave it running. You’ll avoid headaches you don’t need, unless you like complications.
The only way I can see to speed up my iMac even more, is to purchase a large 3-4TB SSD and run it as my only internal drive. That might be a little faster. But as you can see from the test screenshot shown above, my iMac is no slouch right now. And 4TB SSDs are still fairly expensive. It might actually be cheaper (and possibly faster) to get a 2TB SSD and a 512GB NVMe module, and run them together with Fusion Drive, although the overall capacity wouldn’t be the same. Food for thought.
Updated 7/30/21: I’ve since bought that larger 4TB SSD and I am very happy with the performance. Also, the Fusion Drive protocols will NOT allow you to create a Fusion Drive with two SSDs (or one SSD and one NVMe). They require the use of one SSD and one HDD. So, final(?) conclusion on this matter: larger SSDs are more affordable now and getting a nice, big one and running that alone without bothering with Fusion Drive is a viable and preferred option. Otherwise, stick with Fusion Drive.
I’ve worked on a few upgrades to our family computers in recent months: my 2008 MacBook Pro, my mom’s 2007 iMac and my own late-2013 iMac (A1419, EMC 2639). This is the tale I’d like to recount for you now, because it’s something that I’ve had to deal with since last autumn (october of 2018), and I’ve just recently (I hope) finalized the upgrade/repair. There’s a valuable lesson in here for the people at Apple Support, if they’re interested.
It started with my iMac slowing down to a crawl over the course of a few days, back in late 2018. At first I thought it was spyware or a virus. I checked everything: every app, every file, every process. I removed apps, one by one, to see if it would fix the problem. It didn’t. I ended up removing all of the apps but those that came with macOS, and the problem still persisted. I wiped the drive clean and reinstalled the OS, then reinstalled the apps anew, one by one. It was just as bad. I ran hardware tests using Apple’s software and everything checked out. I scoured the web for solutions. There were some suggestions online that iCloud could cause slowdowns when the syncs weren’t going properly. I checked the Apple Server Status page and indeed they’d been having some problems with iCloud, but they were marked as resolved. I checked Photos and there were major issues: my photos weren’t syncing properly across my devices, and there were image compression/corruption (?) issues going on, with diagonal blue lines appearing all over my recent photos, lines that persisted even when opening the photo at full resolution.
I called Apple Support and began a series of interactions that did not end with any sort of solution. From the start, they agreed that iCloud was causing the slowdowns and had me go through a series of steps such as logging in and out of iCloud, disabling and enabling the various iCloud syncing options, etc. to no avail. Mail started acting up as well, so they suggested I disable Mail syncing, because I had “too many messages in my mailbox”. Documents started acting up, with iCloud Drive showing up empty on my computer, so they suggested disabling that. In spite of the fact that I’d already done it, they tried to convince me to reformat my computer and start fresh. I kept getting a hunch that something else was amiss and asked them if they were sure this wasn’t a hardware issue. They said no. They told me to wait for the photos to finish syncing, then enable the other iCloud features one by one, and things would get back to normal. They didn’t. We kept going back and forth, with Apple Support posting one update per day (or less) to the case, asking me to do this and that, and even though I’d complete their requests and post more updates during the day, they wouldn’t respond till the next. I offered to pay a case fee to expedite issues. I explained to them that this was my main computer and I couldn’t do my work. This went on for weeks, with me getting more desperate and the techs telling me they’d stop helping me because I couldn’t stay calm.
With things going nowhere and Apple Support techs who couldn’t care less, I decided to exercise the “nuclear” option. I found Tim Cook’s email address and wrote him an email. I didn’t expect a response, but I wanted to vent. To my surprise, a short while later I got a response, not from him but from his office, promising me my issue would be forwarded onto someone who would get back to me. I waited a couple of days and… nothing. No one contacted me. I figured I’d try my luck again. I got the same response, but someone finally contacted me and connected me directly with an advanced support technician. There’s apparently a “third tier” of tech support that is only available on a case by case basis; I guess after almost a month of my computer being down for the count, I qualified.
It took a while longer to get to the bottom of the problem: daily communications, screen sharing sessions, uploading log files to Apple servers, trying various steps, etc. It took over a week. iCloud was at first to blame, then Adobe software, then finally, after my case was put in front of a senior technician directly responsible for iCloud connectivity, my issue turned out not be software related at all, but caused by hardware. Hold on to your hat, because as it turns out my HDD was going bad. That was it. It was as prosaic as that! In total, I’d lost over one month of my time and I had to appeal to Tim Cook’s office, all because Apple Technicians couldn’t pinpoint a failing hard drive from the get-go.
Once the problem was found out, it was an easy enough fix. I opened up my iMac and replaced the HDD with a fresh one. I also found and installed a piece of software called DriveDx, which gives detailed stats about hard drives and can let you know of a bad drive before it actually fails. Most drive diagnostic apps rely on the S.M.A.R.T. status flags, but that’s not enough. DriveDx does a whole lot more. The app quickly let me know that the blade SSD (the second half of the Fusion Drive on my iMac) was also close to its lifespan. Since it was still working okay, I decided to hold off on replacing it last year, choosing to monitor it with the app and only replace it when it was close to failing. Here is a gallery of photographs from the time I replaced the HDD. I also chose to take apart the chips and heat sinks and to replace the thermal paste, which had become dry and cracked. Before I put my computer back together, I cleaned the case and the parts thoroughly with a brush and soft cloth, because a lot of dust had accumulated inside and on them. After I put my iMac back together, it worked beautifully once again. I know I could have replaced the HDD with an SSD, and I plan to do it in the future. It’s just that I want to get a 3-4 TB SSD and their prices are still a bit high.
A few weeks ago (about seven months after replacing the HDD), DriveDx told me the blade SSD was fairly close to failing, so I chose to replace it with a Samsung 960 Pro M.2 512 GB NVMe SSD module that I already had, so this upgrade only cost me about $16 for a Sintech NGFF M.2 NVMe SSD Adapter Card, which seems to be the card everyone recommends for MacBooks and iMacs.
I’ve enclosed a separate gallery of photographs of this upgrade below. You’ll laugh when you see one of the photos, so I’ll explain: in what seems to be an iMac design flaw, the cylindrical bracket that secures the screw for the blade SSD is only glued to the motherboard. Unlike every other screw bracket which is secured directly to the motherboard with metal, this one is not designed into the motherboard. Underneath it, on the other side of the motherboard, there are circuits running right across that spot. It seems to me like at the time (2013), the blade SSD and its screw bracket were afterthoughts of sorts for the hardware design team. My bracket came right off the motherboard. I had no glue in the house, only some silicone-based adhesive that takes up to 24 hours to harden, so I put a dab of that under it, tightened the screw and stuck a wood shim in-between the case and the top of the screw, to hold it in place while the adhesive hardened. I know it looks terribly unrefined, but it’s been working fine.
There are two things I should tell you about this upgrade: (1) this particular SSD tends to run hot, so DriveDx will warn you about its temperature, and (2) after booting up my iMac for the first time, it didn’t see the new NVMe SSD, so I powered it down, opened up it up again (thank goodness I hadn’t yet closed it completely) and re-seated the SSD and its adapter in the blade SSD slot. After I did that, it saw it, recognized it, and I was able to boot into recovery, go into Terminal and recreate the Fusion Drive, then reinstall the OS and restore my data.
In spite of the temperature warnings, my iMac has been working great so far. I noticed a bit of a speed boost, but since I’m still using Fusion Drive and I’m tied to a spinning hard drive, a lot of the oomph of the SSD can’t be seen. I suppose I could have chosen to install the OS on the SSD and keep my files on the HDD, but I prefer to work without complications. A single 3.5TB volume works for me. If I could have a single 24TB drive that holds my OS and all my files (that are currently sitting on three external hard drives), I’d be happy with that.
I would like to thank Tim Cook’s office for responding to my messages and getting me out of a real bind. I was at the end of my wits at the time, so the tone of my emails to them was gruff and biting; they could have chosen to ignore me. I don’t know what I’d have done if they hadn’t stepped in. But I do wish I didn’t have to exercise that option. Apple Support should have found out the issue from the start. It wasn’t something arcane, it was a simple drive failure.
I bought a Drobo 5D on the 29th of December, 2012, after experiencing catastrophic data loss with the 1st Gen and 2nd Gen Drobo. During multiple phone conversations with Data Robotics’ CEO at the time, Tom Buiocchi, he convinced me that they were much better-engineered than previous-generation Drobos and they had built-in batteries and circuits that would automatically shut them down safely in case of power loss. I was also told the new firmware running inside them would be checking the data constantly to guard against file corruption or data loss. All of these were problems that I’d experienced with my existing Drobos, so even though I was exhausted after my ordeal and so weary of storage technology, I went ahead and purchased the new model and also agreed not to publish an account of what the Drobos had done to my data at the time. I want to make it clear that I paid for my Drobos, so I didn’t feel that I owed him anything, but I did want their company to do well, because back then they were new and deserved a second chance. Now though, there is no excuse for the multiple times their Drobos have lost my precious data. They’ve been around for 13 years and they’ve had plenty of time to make their technology stable.
What’s probably kept them on the market is the willingness of paying customers like me to take a chance on the uniqueness of their proprietary RAID: as far as I know, they are (unfortunately) the only RAID array that lets you store a large amount of data on a single volume that grows automatically as you add drives and also protects (except when it doesn’t) against hard drive failures.
However, after five years of using my Drobo 5D on a daily basis, I can tell you without any doubt that the Drobo 5D does not keep your data safe. Look elsewhere for safe data storage devices. I certainly cannot trust it with my data anymore, so I’ve elected to publish my account of data loss from 2012, as well as an account of my present data loss. Caveat emptor, lest you also lose your data. I’ve also had multiple problems with my two Drobo 5Ns. Because of these problems, some of which have led to significant data loss and to significant time and effort expended in order to restore my data from on-site and off-site backups, I cannot place my trust the new Drobo models that are available, either: I’m talking about the 5D3, 5N2 and the 8D. I see no reason at all to spend more money on more empty promises from Data Robotics.
I can’t say exactly what happened with my Drobo 5D. Drobo Support could not or did not choose to tell me, even though I sent them multiple diagnostic logs from the Drobo and I asked them to tell me what happened. My best guess is that the 5D kept “healing” a 6TB WD drive with bad sectors instead of asking me to replace it. Then a different drive from the array failed. Once the Drobo told me to replace that drive, I did. But during the process of rebuilding the data set, the Drobo 5D decided it didn’t like the 6TB WD drive it had been healing, and chose to tell me that I needed to replace it now. When I did, it would tell me I shouldn’t have taken it out and I should put it back in. I’d put it back in, and then it’d go through its internal processes, only to find out that it wouldn’t like that drive again, but it didn’t stop there. It’d reboot. At first this reboot cycle would take 10-15 minutes, allowing me to copy some data off it, but then it began doing it every 5 minutes. Since it takes a good 3-4 minutes to boot up, this meant I’d have only 1-2 minutes to copy data off before it’d reboot again. This was not workable. After opening a case with Drobo Support, they told me to put it in Read Only mode. You press Ctrl+Opt+Shift+R while you’re in Drobo Dashboard and this reboots it in that mode, which means it’s not going to try and rebuild any data internally, it’s simply going to present the volume to you as it is. This also turned sour quickly, because after allowing me to copy a small amount of data for a few hours, it began that same 5-minute reboot cycle. So I had no way of getting the data off the damned thing unless I mounted it through Disk Warrior and used the Preview app built into that software, which meant having to put up with USB 1.1 transfer speeds. More on the reduced transfer speeds below.
My take on the situation is that it’s a failure in Drobo’s firmware design. It should have asked me to replace the 6TB WD drive instead of working around its bad sectors. Because it didn’t ask me to replace it in time, it then failed when rebuilding its data set after the second drive went bad. That’s not two drives going bad at the same time, that’s a drive going bad and a few weeks later another drive going bad. The Drobo had plenty of time to fix the ongoing situation if its internals had been programmed correctly, but it didn’t, because of inadequate firmware running the device. That’s bad technology at work, causing me repeated data losses.
The 6TB WD drive had 9 bad sectors, but the Drobo 5D kept insisting on healing it until it was too late.
Here’s another example of Drobo’s crappy firmware: for the past three years, I have had to force my iMac not to go to sleep, because every single time I’d wake it up, the Drobo 5D would refuse to mount, forcing me to reboot the iMac and/or the Drobo and also disconnect/connect the Thunderbolt cable from my iMac in order to get the computer to see it. Data Robotics tried to fix this horribly annoying problem (which can also cause data loss) through multiple firmware fixes, but I can safely tell you that they still haven’t fixed it. Before I stopped using my Drobo 5D, I was on Drobo Dashboard 3.3.0 and Drobo 5D firmware 4.1.2, while my iMac is on MacOS Mojave 10.14.2, and this problem still very much occurred. Well, it definitely occurred before my 5D crapped itself. Oh, how I’d like to be back to those simpler times when all I had to deal with was keeping my iMac from going to sleep! But no, now I have to deal with massive data loss, once again. For the goddamned umpteenth time, Drobo!
Except I didn’t cause the disconnect, Drobo’s crappy firmware would cause it to happen, over and over and over…
What do you think would happen after enough improper disconnects? That’s right, volume corruption!
Look at the difference in size between the two directories. The first is a backup, the second is supposed to be the working directory. That’s the Drobo 5D, actively losing your data… Only 54,890 missing photographs? Thanks, Data Robotics! (This was in 2017.)Here’s more proof of data loss from 2017. Those yellow triangles are indications of missing video clips.Here’s proof of more data loss from 2015, when even Disk Warrior couldn’t get my data back.
Data Robotics marketing speak tells you of super-fast transfer speeds and protection against data loss when you buy their devices. The things you most need to remember are that you will experience data loss (that’s a given) and you won’t be recovering your data at super-fast transfer speeds such as USB 3.0/USB-C or Thunderbolt 1/2/3. You will instead be forced to use Disk Warrior’s Preview application (if you’re on a Mac) and you will be recovering your data at USB 1.1 speeds. That’s right, take a moment to think about that! In order to recover data from my Drobo 5D, I have to use Disk Warrior (a Mac app known for its ability to recover failed volumes), because it won’t mount any other way. It certainly doesn’t mount through the Finder or through Disk Utility.
“Work” is the Drobo 5D volume that refuses to mount
The size of my Drobo volume is about 12 TB. Thank God some of it was backed up locally with Resilio Sync and some of it I recovered from online backups with Backblaze, so I only needed to recover part of my data from the Drobo 5D itself, but for a single 3 TB Final Cut Pro library, it took roughly 300 hours to transfer it to another drive! 300 HOURS! It might be somewhat tolerable (in some masochistic sort of way) if I knew the damned thing would stay on for all that time, allowing me to copy the data in one go, but you never know when it’s going to restart. It can go at any time.
I’ve posted illustrative screenshots below. Feel free to do the math yourselves as well. Also think about how much worse this problem gets when your Drobo volume is much, much bigger. The Drobo 5D and 5D3 can go to 64 TB, while the Drobo 8D can go to 128 TB. Do you really want to be stuck copying 50-60 TB of data at USB 1.1 speeds? How about 100 TB of data? Think about that before you click on the Buy button and get one of those shiny black boxes.
The hard truth is that you can’t put all your eggs in one basket. Having all your important data on a single volume, which is what the Drobo lets you do, is a dumb idea. If that volume goes, all your data goes along with it, and it can take MONTHS to recover it from backups.
After more than 12 hours… Transfer speeds vary between 100 kB/s – 4 MB/s but mostly stay around 100 kB-200 kB/sTwo days for 350 GB… Great…
I don’t know about you, but I don’t want to spend anywhere between $5,000-$10,000 with a data recovery firm every few years, when a Drobo unit decides to fail and lose my data. But that’s what Drobo Support advises you to do from the very beginning. They advise against trying to recover the data yourselves, even while working with them, and instead try to convince you to send your Drobo into a drive recovery firm. For a device that’s supposed to protect your data and a company that brags about “protecting what matters”, that’s disgusting. How exactly are they protecting your data when their devices fail miserably every few years? I guess “protecting what matters” really means “protecting their bottom line” by ensuring suckers keep buying their product.
Do you want to know what happens when I try to open a ticket on the Drobo Support website? This:
For months now, all I get is this grey page with a rotating status ball when I visit Drobo Support
If a company can’t even fix its support page so that it loads up for its customers, that’s cause for worry. I tried explaining the issue to them, I even sent them screenshots, but the techs I spoke with seemed unable to comprehend why the website wouldn’t load. My guess is they’ve got a geofence that stops visits from Romanian IP addresses.
When it comes to their marketing people, I cannot describe them as anything but a bunch of slime buckets. Twice now they’ve ignored me in dire situations when I reached out to them for help, hoping they would redirect my messages internally and get someone to pay proper attention to my case. Back in 2012, back when my two DAS Drobo units lost over 30,000 files, they began ignoring me after I told them what had happened. Now, in 2018, they pretended to help, just so it looked good at first glance on social media, but they didn’t follow through on their promise and ignored me afterward. See my comments on their tweet here. I also wrote a message to some guy who calls himself the Drobo CTO but gives no real name, asking him to have a look at my case, but he ignored me. If it’s a fictitious account then it’s understandable, but if it isn’t, then he’s a turd as well for ignoring a legitimate and polite request for help from a customer.
I also reached out to Data Robotics’ current CEO, Mihir Shah, to ask for his help, and after an initial reply that said, “I am looking into this case for you. We will get back to you. Thanks.” — his line went dead. I am left to conclude that he’s of the same breed as his marketing people, who did the same thing when I reached out to them.
You might be asking yourselves why I chose to reach out to these people when I had already opened a case with Drobo Support? Because I felt the technician handling my case wasn’t doing a good job. This case wasn’t a freebie, I paid for it, and I wanted to get real help, not bullshit. Here are a few examples of the kind of “support” I received:
He told me to clone a drive using Data Rescue (a Mac app). This was supposed to fix the situation, but it didn’t, because he told me to clone the wrong drive. That’s right, he had me waste an entire day cloning the wrong drive, before I pointed it out and he came back with sheepish excuses, asking me to clone another.
He let me buy two new 8 TB hard drives in order to go through the data recovery process, without saying a damned thing, before another tech stepped in to tell me I should have bought a 6 TB hard drive instead, because the cloning process needed for my scenario requires a drive of the exact same size as the 6 TB WD drive that the Drobo 5D refused to use anymore. Why couldn’t he tell me that from the start? Besides, isn’t the whole point of BeyondRAID, which is Drobo’s proprietary technology, to let people use drives of varying sizes? There goes that advantage when you need it…
Having worked in IT for a long time, and having worked my way up from the help desk, I could tell the technician didn’t give a crap about the case. It was clear to me that he wasn’t interested in helping me or solving the case, he was simply posting daily case updates of 1-2 sentences with incomplete and unclear replies to my questions on how to proceed and dragging the case on, probably hoping I’d give up.
This time around, I’m faring better when it comes to data recovery than I fared in 2012. Unfortunately, I don’t have a full local backup of my data, even though the Resilio Sync software was supposed to mirror it from my Drobo 5D to my Drobo 5N. I do have a full online backup with Backblaze, but getting back about 12 TB of data through online downloads will prove to be a cumbersome and slow affair. They only let you download up to 500 GB at a time, which means that if I want to download all of it, I’d have to create at least 24 restore jobs on their servers, each of which would generate a ZIP file that would need to be downloaded and then unzipped locally. I could also choose their HDD recovery option which ships your data to you on 3.5TB drives, but I’d need 4 of them, which means it would cost $189*4 or $756. It would probably work out to something like $850-1,000 for me in the end, because I would incur a high shipping cost from the US to Romania and I would also be responsible for customs fees. Theoretically, Backblaze offers a money back guarantee if the drives are returned within 30 days, but I doubt I could return them in that time span, given they’d have to make it to Romania and back. They’re supposed to be working on a European data center and it might open this year, and while that’s going to be nice in the future, it’s not going to work for my situation at this time. No, the much more workable solution would have been for me to have a full local backup of the files on the Drobo 5D. But thanks to Data Robotics, that option got shot down and I didn’t even find out until it was too late…
You see, the Resilio Sync software was set to mirror my data from the Drobo 5D to a Drobo 5N. On my gigabit network, that would have worked just fine. The app was running on my iMac and it was also running on the 5N through the DroboApps platform. The software had a few months to do a proper job syncing my data between the two devices and it indicated to me that it had done it. Unfortunately, after my Drobo crashed, I found out that it hadn’t. The reasons offered by Resilio Support were the following:
The Drobo 5N’s processor wasn’t powerful enough
The Drobo 5N’s RAM wasn’t enough
The Drobo 5N ended up using swap memory and somehow it messed up the sync logs, which caused it to think it had finished the sync when it hadn’t
If I’m to take them at their word, the Drobo 5N is too underpowered to handle Resilio Sync. That’s not to say there isn’t plenty of blame to be assigned to the Resilio Sync software as well. After all, it should have kept track of the data sync accurately. What good are they if they can’t perform at their main advertised task?
If I am to look at what Drobo is saying about the successor to the 5N, the 5N2, namely that it “provides up to 2x performance boost with an upgraded processor and port bonding option”, that would certainly mean the 1st gen 5N is underpowered. Finally, who do you think recommended Resilio Sync to me? It was none other than Data Robotics, whom I called to ask for details on DroboDR, their own data sync application. They said it was too barebones of an app for my needs and suggested I look into Resilio Sync, which was much more robust and full-featured. Thanks once again to my trust in the Drobo, my backup plans got sabotaged, forcing me to waste weeks of my time recovering data from a failed Drobo volume.
But enough about the Drobo 5N, I’ll have a separate post where I’ll talk about how that device has also lost all of my data recently… Back to the Drobo 5D.
To recap, the Drobo 5D lost my data in 2015, 2017 and 2018. That’s three separate, significant, data loss events, the last of which wasted more than a month of time till I was back on my feet.
I am back on my feet now. It’s the 26th of January 2019 and I finally got the last of my data recovered. I’ve been down, unable to do my work, since the beginning of December 2018. I can’t even remember what day it was the Drobo 5D failed. It’s a blur now.
I ended up buying a third new hard drive, specifically a 6TB hard drive, to match the size of the drive that the Drobo didn’t want to use any more, and I cloned that drive, which turned out to have 5 clusters of bad sectors, onto the new drive, using Data Rescue software for the Mac. I had to use its “segment” cloning mode, which attempts to get around the bad sectors, and this meant the cloning operation took more than 120 hours!
A screenshot taken after 117 hours…
Around hour 122, my iMac crashed and I had to reboot it, so I don’t know if the cloning operation was completed or not. My screen went black and stayed that way. It could be that the cloning ended and then my iMac crashed, or simply that my iMac had had enough of staying up to work through a lengthy cloning process and went poof. I don’t know, and I wasn’t about to start cloning the damn drive again. I took my Drobo 5D out of Read-Only Mode, stuck the cloned drive into the Drobo 5D and booted it up, expecting some data loss. Besides, even if the cloning operation had completed successfully, I’d have still lost some data. If you look above, you’ll see that there were 298,2 MB of bad, unrecovered data. That wasn’t from one portion of the drive, that was from five different portions, because there were five significant slow-downs in the cloning process.
The 5D booted up and started its data protection protocol
Once it booted up, I started copying my data off it, not knowing how long it might stay on. For all I knew, it could go into a vicious reboot cycle at any moment. I’d already recovered all I could from Backblaze and from my local backups, and I needed about 6-7 TB of data from the 5D before I was back up to normal… “normal” being a loose term given how unreliable the Drobo is and how much data I’d already lost.
Sure enough, data loss soon reared its ugly head. Thanks, Drobo, you bunch of slime buckets!
One of the damaged FCPX libraries with lots of missing video clips
So far, two FCPX libraries containing the videos I’d made in 2017 and 2018, are damaged. I don’t know if I’ll be able to get those videos back.
I consoled myself with the idea that at least I was able to recover most of my data at Thunderbolt speeds, not at USB 1.1 speeds, so that saved me about 2-3 more weeks of painful waiting.
All of my data is now off the Drobo 5D, and I don’t plan on using it again. I’ve put it in a storage closet. I’m done with it, and I recommend you also stop using your Drobos, if you are using one, because it’s only a matter of time before they will lose your data. In my opinion, they offer no more protection than using individual drives. They just cost more, force you to buy more hard drives and they make more noise. My office is so much quieter now that the Drobo 5D is gone. I’m using individual drives which I plan on cloning locally. I will continue to use Backblaze for cloud backups, and I may also do some periodic local backups to a NAS of some sort. I have found out during these past few months that I also cannot rely on the Drobo 5N NAS to store my data, because one of my Drobo 5N units has just lost all of the data I’d entrusted to it. I am now in the process of recovering that… It’s non-stop torture, time and data loss with a Drobo!
I have been a Drobo customer since December 2007, when I bought my first Drobo. I was also among the first Drobo “Evangelists”, as they called their enthusiastic customers back then. It’s now January 2019, and I am done with Drobo. I was a loyal customer for 12 years and I stuck with them through an incredibly disheartening amount of data loss, problematic units and buggy firmware. That’s enough of that. Caveat emptor. Drobo no more.
I’ve had multiple Drobo units since 2007. To this day, I still enjoy adding a hard drive to a Drobo. It’s one of those things that can be an ordeal on other tech, but on a Drobo, it’s been made fun through proper planning and design.
It lets you that it’s low on space, you order a drive, and when it comes, you look at the app, which tells you exactly what size-drive is in each bay. Pressing a small lever on the side of the bay releases the drive, which slides out. You put the new one in, the Drobo immediately checks it and formats it, then begins striping the data set across it. By the way, that’s a screen shot showing my Drobo 5D.
I love this process. It’s so simple and so fun! The Drobo doesn’t care what hard drive you buy, as long as it’s larger than what you already had. It allows you to grow the capacity of your Drobo in time, as the prices for newer, bigger hard drives decrease, without any sort of headaches. This is technology done right.
Raoul Pop 