Not too surprising, they killed their server hardware line (Xserve) in 2004, and the Mac Minis that people were using as substitutes lost their quad-core option in 2012.
The number of people hanging on to Macs as servers has to be vanishingly small, and Apple doesn't do small markets anymore. There are currently rumors of a "Pro" focused Mac Mini, yes, but that's going to be comparatively expensive and I doubt it could win back much server market even if Apple wanted to.
They'll stick around as servers for iOS builds because there's no other option (and the Pro Mini would be nice for that), but hanging on to Macs for any other sort of server outside of a home file server is a lost cause.
they killed their server hardware line (Xserve) in 2004
Minor nit: 2011 [0] But similar to you, I'm surprised macOS Server is even a thing anymore. What are you supposed to run it on? That six year old quad-core Mac Mini? The oh-so-long-in-the-tooth trash can Pro your department still clings to? Off the top of my head, I'm a little stumped as to what I'd use to best utilize Xcode's distributed build functionality. Apple hasn't made what I feel is an appropriate tool for that job in at least six years.
The only logical "server-ish" thing I can think of is a build server for projects requiring Apple-only dev tools. I doubt anyone is using OS X server for that though. I've seen Mac Minis used for this, and I recently learned it's not that hard to run macOS in a VM, but the latter is probably not in keeping with the EULA.
What if Apple offered a configurable "hardware blessing" license for sale (perhaps tied to the purchaser and a specific macOS version, not the hardware, and non-transferable)? It would allow you to legally run x copies of macOS 10.y on non-Apple hardware (either directly or as a VM). If Apple did that, I'm sure our IT department would jump on it. They're very frustrated about having to maintain Mac Minis just to host macOS VMs.
Failing that, what exactly constitutes Apple hardware, legally? Could I run a macOS VM on a big Dell server if I zip-tie a Mac Mini motherboard to the enclosure?
It bears reminding that the EULA is not enforceable in all jurisdictions. I am not suggesting copyright violation - just that running a legit macOS in a VM on a non Apple hardware is sometimes very handy. For instance for iOS building or for quickly regression test you Mac app across a handful of macOS versions.
That’s a really good question. You know how we always said there’s hope that Apple won’t ditch the Mac since they themselves couldn’t stand to use Windows or Linux for their workstations...
I have no knowledge of their internal setup, but I suspect they have a big room full of racked ESX servers (non-Apple hardware) running builds on macOS VMs. And probably a ton of networked iPhone motherboards running automated tests.
I just wish they would make that kind of setup (macOS VMs on non-Apple hardware) legal for other developers.
My understanding is that internally when they have tasks that their own hardware isn't appropriate for they use a combination of Linux and macOS running on commodity PC hardware. They have an internal version of macOS for the commodity PCs. The person who told me hinted they got machines from a big name vendor like Dell or HP. I was informed of this in a discussion about AR and VR. Basically we all asked how Apple even did work on this internally when their own machines clearly had some issues with that sort of workload. This was like six months ago and related to speculation about showing the new Mac Pro at WWDC 2018, which didn't end up happening.
Interesting. I wished Apple had a Mac Pro that is basically the old Cheese Grater, and rack based in the size of 3U or 4U. It should aim to hit both market with a single design. I mean even if they do it for internal usage, along with other niche for Server and Pro uses would be enough to sustain the business.
When I worked as a consultant for Apple Retail Software Engineering, we did have huge clusters of machines running VMware — on the biggest, beefiest, Xserve hardware that could be bought. Maxed out on RAM, CPU, and everything else that could be crammed into those boxes.
I was told that the entire final run of hardware was reserved for Apple internal use.
Our Jenkins CI/CD server was just a VM on the cluster, and ran Xcode to do the iOS and MacOS X builds on one set of slaves, and the Linux back-end server software on another slave.
> I was told that the entire final run of hardware was reserved for Apple internal use.
I get the sense that Apple's strict devotion to its product line is (to put it nicely) creating a discontinuity between their own practices and what they offer to others. It makes me think Apple needs a separate company or brand to offer practical utilities (and perhaps consulting) to companies wishing to deploy Apple software -- things like licenses to use macOS on non-Apple hardware (like a Dell EMC VxRack SDDC[0]), productizing some of their own IT solutions, special iPhones with a 1Gb RJ45 port attached for automated testing, and even "catch the vision"-style marketing to promote their idea of a data/compute center that integrates macOS and other Apple technologies.
I've heard for major things like the OS, they have standard server racks that I'm guessing have an equivalent of hackintosh to enable building macos / ios with it.
I’ve got a Mid 2010 Mac Mini running regular macOS I use as a light duty home server. Went to update it to Mojave this weekend to find Apple dropped support for it! F-me for having still-working hardware, I guess :(
The only thing really is Open Directory for centralized accounts, and you only need a Mac Mini for that. We do this in an isolated secure network where we keep certain assets for customers who have confidential data. For everything else, we moved to Windows years ago.
I suspect many people (myself included) were using Server.app solely for one of the services (File Server, Caching Server, Time Machine Server) they since moved into the standard macOS install.
Why do they even bother making macOS server anymore? They clearly don't care about the corporate enterprise. They keep stripping away features. No DNS, no DHCP, no VPN, what is anyone gaining using macOS server? If you use it you likely are going to have to hack together a bunch of services just to get it functioning.
The announcement sounded to me like they're refocusing Server.app for Profile Manager / MDM than one-application-do-it-all approach like they did previously. I think this benefit even more for enterprise users, because they're likely not going to use the bundled Apache/Dovecot/etc. that came with Server.app, but rather using existing infrastructure they already had in place anyway.
Edit: one thing to note: macOS Server is not an entire OS, but rather a Server.app application running on top of existing macOS installation that you can download from Mac App Store.
Profile manager / MDM on MacOS Server is a goddamn nightmare.
It's amazing when it works, but when it breaks you're SOL. Case in point is I inherited a semi-broken system when I started my job 3 years ago. ~40 enrolled laptops but couldn't add any more. Peeking into the logs the SQL server built into the .app bundle broke/corrupted.
Many others in the Apple forums had the same issue but no resolution and the solution ended up being "flatten and reinstall".
For a small business, sure Profile Manager might work, but scale beyond 10-20 users Jamf Pro is the way to go.
One of my friends spoke to an Apple engineer about problems he was having "scaling" Profile manager. The engineer basically laughed and said that they never actually intended for anybody to use Profile Manager.
1) It probably makes them at least a noticeable percentage of revenue, if not much, so that cutting it out completely would maybe cut funding for a dozen employees. They'll need to dwindle sales down first, probably by deprecating it, so that they can justify the cut internally.
2) Even if they're backpedalling on the idea of macOS as a server and pushing for migrating to open source tools, they don't want to openly admit this, that always looks bad and unprofessional.
3) They're probably phasing it out slowly and in the next release or two they will probably deprecate it, and remove it completely from MAS in the next major macOS update.
4) They'll want to wait until they can play up their contribution and downplays that they were on the wrong side of history: "We're deprecating macOS Server, which has been a boon and a huge innovation in this space for 20 years, and helped $N millions of customers, but at this point the open source community has finally caught up to our quality."
5) At that point, the remaining useful features will be integrated into macOS proper for free, and they'll tout how they're benefiting everyone by doing this and simplifying everyone's workflow and saving everyone money.
> ) Even if they're backpedalling on the idea of macOS as a server and pushing for migrating to open source tools, they don't want to openly admit this, that always looks bad and unprofessional.
I think it looks way worse to just let it limp along and slowly die. Fail fast, learn your lesson, and move on.
>1) It probably makes them at least a noticeable percentage of revenue, if not much, so that cutting it out completely would maybe cut funding for a dozen employees.
They are APPLE, they have ridiculous cash. If said employees are of a high quality/standard then I am sure they can figure out something else for them to do at Apple. Even if it means paying them for a few months while they get trained up on a new project/department.
> I think it looks way worse to just let it limp along and slowly die. Fail fast, learn your lesson, and move on.
The benefit of limping along and dying slowly is that it can be relatively quiet and uneventful, and at least the project dies with the appearance of dignity, even though it's actually the opposite. Investors applaud "sticking it out" even though the engineers knew it was dead for years. Review sites write poignant eulogies, instead of writing angry rants which they would do if it was cut short.
The thing about software (and probably most kinds of products) is that there's always someone who relies on it, no matter how dumb of an idea the product is, and usually that group is larger than you expect or seems reasonable. Those people are often the ones who yell the loudest when you inconvenience them, which is why Apple has to pander to them by torturing macOS Server to death instead of euthanizing it.
> They are APPLE, they have ridiculous cash.
They have always acted like a start-up that's strapped for cash. I've known many Apple employees and it's just been part of the culture, that if you work at Apple, you're going to wear many hats and they're going to utilize every second of yours to 200%. It was especially so this in the days of Jobs, and I doubt it's changed much in the past decade.
I am an investor (not in Apple, but other companies). Believe me, as an investor, there is nothing more that I want than for a company fail fast, and drop things quickly when they aren't working out. The worst thing for an investor is when a company tries to "stick it out" (investors read that as "wasting money").
> I think it looks way worse to just let it limp along and slowly die. Fail fast, learn your lesson, and move on.
I'm sure Apple are well aware of the lesson and are simply doing the right thing for the folks who did invest in macOS as a server. It is possible for a company to learn their lesson and do the best thing they can for their customer. Apple's financial position enables them to give a much longer runway than a typical startup changing direction.
>doing the right thing for the folks who did invest in macOS as a server.
The "right thing" at this point would be to help you clients move to something else (jumpcloud? AD, etc.) instead of continually putting out a diminished version of your product.
That's only one way to look at it. There are lots of weird integrations and third-party products that OS vendors have to account for in a decision like this. I don't know how they set dates for end-of-life on their products, but many vendors are obligated to provide support longer than a reasonable person might expect.
> No DNS, no DHCP, no VPN, what is anyone gaining using macOS server
Probably because the big enterprise go for Microsoft and Active Directory, and the small ones go for nothing, just a "home" router, and iCloud/GApps/O365. Or you can spin up a Linux box and do most of this if you start getting bigger. The only place I can see this is large Mac heavy shops like designers etc. But even then, most of the MDM stuff can be hosted outside.
Yep, you're right. They've been slowly decoupling services from macOS Server to macOS. A big one is content caching which was added to macOS High Sierra under System Preferences > Sharing > Content Caching. It will cache iCloud and software updates so that devices checking for updates go though that cache.
None of the macOS server functionality is particularly relevant to build servers. We don't use it at all despite having a decent-sized build fleet and instead do everything that doesn't have to run on macOS on linux servers.
Looks like the original title was something like "macOS Server gets worse in 5.7.1" but was then changed to the page's official title, "Prepare for changes..."
This is a case where I'd argue deviating from the "official" article title is warranted, as it more clearly explains the aim behind the submission, or to what specifically the submitter wishes to draw our attention.
I agree that at times it might be slightly counterproductive but then again if you start making exceptions it's going to be a mess every time as we always argue back and forth whether a particular submission is acceptable or not.
I would also argue that hiding "the aim behind the submission" is soft of the point, submissions should be able to stand on their own. If more context is required then a better link should be found IMO, a HN submission title is not the place for an opinion piece. Worst case scenario add a comment in your submission explaining your rationale.
>I would also argue that hiding "the aim behind the submission" is soft of the point, submissions should be able to stand on their own. If more context is required then a better link should be found IMO, a HN submission title is not the place for an opinion piece. Worst case scenario add a comment in your submission explaining your rationale.
I feel like that then runs afoul of HN's other guideline to prefer primary sources.
Consider the following scenario: PopularSoftware v.6.2 sneaks in a one-line mention in its changelog that the software will now proactively scan the contents of your computer. This would obviously be newsworthy and would make for a popular point of discussion for HN. How best should that be communicated in the submission? If we follow HN guidelines, the submission title should be copied from the main page, "PopularSoftware v.6.2 is released!" That doesn't really make it clear to the HN user casually scanning the front page that the submission is really trying to draw your attention to that particular, controversial bullet point.
An alternative would be promoting someone's blogpost or analysis of PopularSoftware 6.2 and its disclosure of hard-disk scanning. But as I mentioned earlier, you'd be promoting a secondary source over a primary source simply for the sake of clarity in the submission title.
That's a fair point but in my experience in these situations the controversial bullet point would be pointed out by the top upvoted submission when you open the comments so it's only one click away anyway. I hope that people on HN don't upvote/downvote stories based solely on their titles, and if I see an upvoted story whose title is "PopularSoftware v.6.2 is released!" I'm definitely going to check what the people have to say in the comments.
I mean you've been here for quite a while judging by your profile, do you really find this to be a practical problem while browsing HN? Personally it's really a non-issue and the potential slippery slope of letting people come up with custom submission titles is not really worth it IMO.
> I hope that people on HN don't upvote/downvote stories based solely on their titles, and if I see an upvoted story whose title is "PopularSoftware v.6.2 is released!" I'm definitely going to check what the people have to say in the comments.
Even if they don't, if the title is too boring they might let the story pass without doing either. A story titled simply "PopularSoftware v6.2.1 released" (it's even a x.y.1 release, which commonly means only small bug fixes) might not attract enough attention to get more than a few upvotes.
> This is a case where I'd argue deviating from the "official" article title is warranted
Not really. "macOS Server gets worse in 5.7.1" is a purely subjective opinion and there's no reason this personal opinion should override the actual article's title.
I personally have never used macOS server and from that context, the actual title ("Prepare for changes to macOS Server 5.7.1") makes a lot more sense.
A term cannot be accurate. I understand you're trying to suggest that "censorship" is an accurate way to describe this particular scenario, but that's your own subjective opinion, which IMO, is incorrect. Censorship necessarily implies an intentional effort to squelch a particular person or idea.
If changing the title doesn’t result in actively squelching the old title, I’m not sure how to interpret this. There’s fundamentally a silent middleman choosing the content on the site.
Is this bad? No. This is the whole positive and negative to active moderation.
> If changing the title doesn’t result in actively squelching the old title, I’m not sure how to interpret this.
"Changing a title" is not inherently an act of censorship because there are many reasons why a title could be changed. Changing a title might be censorship, but not always. In this case, it is NOT censorship because the title was modified based on a pre-established submission rule (i.e. from the HN guidelines "use the original title, unless it is misleading or linkbait; don't editorialize."), NOT due to any particular prejudice against the ideas expressed in the title.
It also seems a little strange to me that changing a title to reflect the author's actual title can somehow be construed as censorship. It seems to me that the person editorializing the title is more aptly described as a censor than the mechanism that restores the title to the original form and cadence intended by the author.
Whether fair or not, i started arguing that apple wasn't a computer company years ago. They are a phone company today and everything that looks or works like a traditional machine, servers included, are a distraction to that enterprise.
As a once-proud and fond apple fanboy, i am biased here. I miss having compelling laptop and desktop options in the Mac space. In the same way that i am still using an '09 pro and a '12 air, i bet some IT folks still cling to their server infrastructure hoping that the glory days return. C'est la vie.
It might have been a big hint when they dropped “Computer” from their name.
But more importantly, Apple has always been a personal computer company. Not an enterprise computer company. What’s more personal than a phone running a real computer operating system in your pocket? More people (ie the “personal”) buy iPhones + iPads in one year than Apple has sold traditional computers in the last decade. If not, the numbers should be close.
In the early 2000s they made a HUGE push for enterprise.
My company I worked at Part Time went all in. It was a disaster for 3 months. The IT department didn't make the decision but the Executive Board.
Just this summer, I now work at another company, and they went 100% Apple (Because students were issued iPads so why not switch everyone over of course). They ended up throwing away all their 1 to 3 year old Mac All in Ones, in all the computer labs, into the garbage (At Apple's Request Around 400 of them) and now students don't have access to a computer and only issued a tablet.
So adults have Mac Laptops and kids have iPads (K-12) seems so strange. I think that this is where Apple would love to put Enterprise with everyone without a special case just uses iPads.
> They ended up throwing away all their 1 to 3 year old Mac All in Ones, in all the computer labs, into the garbage (At Apple's Request Around 400 of them) and now students don't have access to a computer and only issued a tablet.
They were in perfectly good working order. I really think this is just Apple's practice. Apple All In One users really don't need the latest and I bet they are just seeing this as something that would cut into their profits.
It’s a happy accident that bring your own device bought Apple into the enterprise and now they are able to leverage that and do partnership deals with the likes of IBM and SAP (?)
I know that you're underscoring that the iPhone is their primary product. However, even though "phone" is in its name and was it's primarily replaced device, the iPhone to me has long since passed being a phone. Comparing time spent on device to the amount of minutes used as a phone, the "dialer" app is one of the least used apps on my phone. Essentially, it's a computer that fits in a pocket to the point we've been seeing designs to power a laptop by sliding a smartphone into a laptop style chassis (maybe not from Apple).
I'm curious what Apple uses for datacenter level servers. I know they've got a big footprint (not MS/Amz/Goog/FB level, but big). I can't imagine them wanting to run a commodity OS and standard tooling, given where they were with WebObjects and what they're doing on the client side.
I don't think they're running a pile of mac pros. Do they run x86 rackmounts with a custom OS? Linux? Windows on azure?
Who says they can't have x86 servers with macOS? Just because they don't sell anything like that doesn't mean that macOS doesn't support it for internal usecases (possibly via a separate in-house build).
> Who says they can't have x86 servers with macOS? Just because they don't sell anything like that doesn't mean that macOS doesn't support it for internal usecases (possibly via a separate in-house build).
Seems insane to spend the money to develop something like that, but not sell it.
It's not that difficult to make MacOS run on selected non-Apple hardware or in VMs for people outside Apple, so it likely has next to no cost for Apple.
It's a little tricky to upgrade the CPUs in the Xserve as they are special versions with narrower lids. You can't use stock Xeons, you have to delid them yourself. It can be done though.
On a sidenote I always wanted to buy an old Xserve just because for some reason they look really good. Not sure they are good for anything tho
I've been tempted, too. Mostly for the nostalgia kick.
I've seen them on fleaBay at really reasonable prices. It looks like .edu, .gov, and .orgs are phasing them out and often a listing shows a stack of 20 or 50 of them, and you can pick the one you want.
They might look pretty, but they're annoying - the drive bays pop out when you press them, which is annoying if you knock the front of one by accident.
I'm really surprised it still exists. I've actually tried to use a few of these services on OS X server at various points in the past. It never worked very well behind NAT; anyone that can afford their own IP space is not going to need hand held through setting up Dovecot or BIND.
The GUI is supposed to make it easy to get this kind of stuff set up without really understanding what it does. But, something strange always went wrong which required trying to understand what the hell the open source software was doing at the same time as trying to understand where in the configuration file hierarchy the apple GUI was inserting its config and what this was doing.
Having tinkered a bit with various services on Linux since, its much easier to cut out the middle man, install from scratch and configure a vanilla installation using the config files. All the helpful comments on the options in the config files make it much easier to understand what's happening anyway.
I tried it years ago and it was a terrible product for exactly the reasons you mentioned - one click smiley-GUI setup that consistently failed to work.
And lots of stuck-on features that I couldn't imagine anyone using in a small office, like a barebones custom Wiki.
I never understood where Apple were going with it. As I understand it, it was a solid product that did what it was supposed to - i.e. help network admins manage a network full of Macs and i-devices.
Then those features disappeared, and it turned into My Little Server, but worse.
Which is a shame, because there may be a market for a good server product, and Apple could have owned it.
From my personal experience macOS Server is not a serious product, nothing works out of the box and everything requires serious terminal tweaking. Will not give it another try
Was excited to hear about this since I imagine a very large number of users had to install the server for update caching and time machine server and now they don't how have to. Since the title would probably be better suited to say High Sierra Gains macOS Server Features.
I don't understand ... they've eliminated the Mail server? or you just have to click a button to install it specifically? or you have to manually install one of the open-source mail servers yourself? (in which case why use MacOS server at all, not say, Debian)
My main issue is with Push email. With OSX Server, you could get an APNS certificate that enabled you to have mail instantly delivered to your iPhone, push for Mail.app with one click. Will there be a workaround for this? I know there's a dovecot version for freebsd that can use the push certificate, but the only way to obtain it was using OSX Server's mail server... Does anyone know if a regular, iOS app push certificate works with this?
How about "Apple stops shipping out-of-date versions of open sourced server daemons (for which they had created a minimal GUI)"
Not that having a minimal GUI isn't nice for users with minimal technical skills, provided that you keep up with current versions.
However, the Server "version" of OSX has long been nothing more than an application that installs a version of several open sourced server daemons along with a enough of a GUI to configure the basics.
Most of my own experience with this has been as a developer/user. The user management is addon software tethered to ActiveDirectory. I've worked in one environment where it worked pretty well, and another where it sucked.
As to pre-installed software, not sure. I always just installed brew, and did what I needed myself. CA was already on the box and scheduled OS updates were, lets just say very rare. (don't do it yourself, it can break the authentication software integration).
First of all you get an MDM service. Doesn’t matter which one but these days this is essentially a requirement. Some MDM’s with good reputation in macadmins community are AirWatch, Jamf, SimpleMDM, MicroMDM (open source).
You then add a couple of more components to get some more enterprise features: Munki for software distribution (free and open source), Chef or Puppet for configuration management (both free and open source) and osquery for monitoring and visibility (free and open source).
Jumpcloud can help with those things, we're an entirely cloud based directory and our system agent is OS agnostic. We also integrate with Jamf and other MDM solutions if you need more complexity.
Disclosure: Jumpcloud employee. My email is in my profile if you want to get in touch.
Is there an iOS MDM provider which will allow you to configure per-app VPNs that connect to an open-source VPN? Microsoft Intune only supports specific corporate VPNs, is that because per-app VPNs require a corporate VPN app that is installed on the phone? Why can't the native IPSEC client perform this function on a per-app basis? It would allow segmentation of Facebook from normal traffic.
This makes it sound to me like we should acquiesce with "cloud" services, but that seems like a very different perspective than their recent privacy note:
> We design our products to limit the collection and use of data, use on-device processing whenever possible, and provide transparency and control over how information is shared.
I never ran Mac Server (at least, not for anything real) but it was always a selling point to me that I could easily run these services myself if I wanted. Including more features in vanilla macOS is great, while dropping other significant features sours the whole platform for me a tiny bit. People (and companies) always want extra power, just in case, even if 99% of them never use it.
>This makes it sound to me like we should acquiesce with "cloud" services, but that seems like a very different perspective than their recent privacy note
Apple is all about the cloud these days. Their privacy efforts are about keeping your data safe, be it in the (their) cloud or on your device.
Sadly trust is not a renewable resource, and i lost mine years ago in cloud providers ability to keep data safe.
As far as content caching goes, making it built in to the operation system is a win-win situation. It lessens load on Apple data centers and speeds up downloads/reduces bandwidth for customers with multiple Apple devices.
The current iteration, however, needs some work. On my old Mac Mini, it is constantly complaining that it needs more space. This will become annoying and I'm sure that some people will find this message anxiety-inducing. For me, I have been meaning to find out whether it stops working or if it purges old content when this happens. I hope it is the latter!
It's a lot easier to manage now. If you just need those features, you no longer need to install macOS server. Just toggle a switch in the Sharing preferences.
Our SMB design office is media-heavy, with lots of very large raw photo shoots and source video files. We've been using Apple Server for file sharing for a long time (running on Mac minis). All other services have been migrated or abandoned. At some point we need to switch to a Windows or Linux SMB server, cloud storage is still cost-prohibitive with the file sizes we're throwing around.
edit: File sharing was on Server.app, now it's just plain macOS-flavored SMB
Been really happy with mine... still running a DS-409+ with 4x 4TB drives. Will probably upgrade or homebrew a freenas box again. Still sour from my last freenas experience in 2012 though.
"Services Migrated from macOS Server to macOS High Sierra"
So I guess High Sierra and Mojave? Though I don't see the Time Machine Server feature on 10.14.0 at System Preferences > Sharing.
I'm using a Time Machine server on Linux (Netatalk), but I suffer from an infinite loop somehow. It never finishes the upload for some reason, and keeps incrementing the amount of MB backed up and still needed to back up.
You have to add the disk as a shared folder in the File Sharing section, then right-click and in Options there's a checkbox to make it a Time Machine share.
It's a bummer, at one point Server was actually a pretty cool product, Network Homes and such were a great feature for Macs and it was approachable for a lot SMB applications. A nice GUI on some good open source features isn't a bad thing. I feel that Apple missed a bit of an opportunity there by being stuck in the 90s and early 00s mold where Macs weren't really considered at all, and thus of course businesses weren't so interested either very often. But times changed and if Apple had kept with it they could have both capitalized that and potentially carved out a profitable niche, it would have lined up well with current worries about over centralization and dependence on the cloud and services if Apple had a friendly product that let people do stuff at home. There were a lot more interesting possible features there that never got and now will probably never be explored in the same way.
Granted, the same can be said about a lot of Apple product possibilities, including desktop Macs period. And there are always tradeoffs: Apple's startup-type organizational structure makes multi-tasking hard, but also makes for fantastic focused products. If they were more traditionally structured they probably would have maintained the Mac a lot better, but would the iPhone have come to be in the same way?
At any rate it leaves a small hole for me at least, I don't mind handling things normally at work but I enjoyed playing with a nice GUI at home, and some small business I know still use Server. I asked for some suggestions on the Ars forums, and for anyone else wondering, one I was given that now looks really promising after first tests is Synology's DiskStation Manager, run standalone if you buy their equipment but also possibly under ESXi vis Xpenology (alongside macOS if that's still desirable for some of the remaining MDM functionality and such). I haven't dug deep into it yet but the GUI and power looks pretty nice, and ESXi will even run on Mac Pros in which case a macOS guest shouldn't even be a license violation. Something to consider anyway as a replacement. For simple network needs those running UniFi kit might find the controller is also enough, though right now it's a little shallow on some core features. Might still be good enough and again a decent GUI.
Anyone on HN have any other ideas for equivalents, with sane GUI defaults even if some CLI usage is needed as well? Are some free general distros these days fine for that, or have popular GUI managers that can go on top? I could see more specific distributions like FreeNAS being good for some applications depending on what exactly Server was being used for. Another storage focused option might be NexentaStor CE, though that is limited to 10TB.
Hmm Time Machine Server. That would come in handy for backing up my laptop automatically to my desktop instead of plugging in a usb hard drive once in a while.
I have been using it for years with great success. I've only had a couple minor problems, which were small in nature and more due to Time Machine itself rather than the fact a server is hosting the service.
(Unfortunately I no longer remember the exact problems, but they were not nearly bad enough to deter me from continuing to use it. I have about half a dozen Mac devices in my home, so the net positive of continuous wireless Time Machine backups has far outweighed said small issues.)
I've been using it with netatalk for a while, recently migrated it to a 2012 MacPro running 10.14. Seems to work about as well as it did on my linux box.
Yes, it's new to me because I never bothered to check out Mac OS server. Now that they moved it into the main Mac OS, I've set it up for one of my laptops and it seems to back up invisibly... except for the first backup where I had to leave the laptop on overnight because it took 7 hours ;)
We wanted to switch to a Linux-based NAS. There was a major issue though: EXT4 does not support long enough directory / file names (or paths). As a result of our extensive folder structure and file naming scheme, we ran into all sorts of issues. In the end, we made the jump to Google Drive where security is probably better anyway in comparison to a local NAS.
I don't think there's a path limit in ext4, but looks like individual file/folders are limited to 255 bytes, while hfs+ (on Linux) is 255 utf-16 characters, which is potentially longer.
The 255 bytes limit seems common (xfs, zfs)-while reiser allows 4kb or so.
What kind of naming scheme do you have that you had a problem? Long, multibyte file names?
IIRC, there's a limit that the whole path passed to system calls has to fit into a single page. For most architectures, that's 4096 bytes including the terminating nul character. (This limit arises because the kernel has to copy the passed path to kernel space before doing any work on it, and allocating more than a single page is costlier.)
Just tried, there's definitely a 255 sigle file/folder name limit with zfs - but I had no problem making very deep folder trees. However 'touch' wouldn't accept a path that was too long - 4kb looks about right.
Bash cd wouldn't let mee leap all the way down in one go - but if I got down there (while ls a * ;do cd a * ;done) - I could create a file (ed: view from "top"):
find -iname ok|wc
1 1 12805
I guess I've got a hard time imagining 4kb path and 255 byte path segments being a serious issue - but maybe my imagination is the problem?
What were you migrating from? Most common filesystems have 255 byte limit for filenames. It looks like most of them don't have a explicit limit on path length, but Linux has a limit of 4096. Any idea how Google Drive is a solution since many of these limits seem to exist on the OS side of things?
The number of people hanging on to Macs as servers has to be vanishingly small, and Apple doesn't do small markets anymore. There are currently rumors of a "Pro" focused Mac Mini, yes, but that's going to be comparatively expensive and I doubt it could win back much server market even if Apple wanted to.
They'll stick around as servers for iOS builds because there's no other option (and the Pro Mini would be nice for that), but hanging on to Macs for any other sort of server outside of a home file server is a lost cause.