MAC Mac Mini as server - Xserve is dead

[Cortical]

Xserve is dead:

http://images.apple.com/xserve/pdf/L422277A_Xserve_Guide.pdf

Xserve Transition Guide

November 2010

The choice now is either Mac Pro or Mac Mini

an extract:

<<

Transition to Mac mini with Snow Leopard Server

...

The June 2010 update of Mac mini improves its server configuration with an all-new enclosure with easy memory access, a built-in power supply, a thinner profile that fits into a 1U space, a faster dual-core processor, faster 7200-rpm hard drives, and DDR3 RAM support up to 8GB. For small/medium business (SMB) and workgroups up to 50 users, a single Mac mini running Snow Leopard Server can typically handle an entire workgroup’s server needs. Workgroups with a larger number of users can consider breaking up server services across multiple Mac mini servers.

While not a rack-optimized form factor, the thinner aluminum profile of the Mac mini fits in a 1U space. Two Mac mini systems can sit side by side on a shelf in a 1U. Several different kinds of rackmount brackets are available from third-party vendors.

Workload guidance

Mac mini is a capable workgroup server able to provide a full set of typical workgroup services—file and print, web and wiki, instant messaging, contact and calendar sharing, and VPN—for up to 50 simultaneous users, or 25 users if Time Machine backup is included. As a single-task server, Mac mini can provide service to a larger number of users.

Performance versus Xserve

Mac mini is designed to deliver services to a workgroup of up to 50 people, or provide a single service to a larger client load. As such, it does not deliver the range of perfor- mance that Xserve does. Customers with high-performance or high-capacity storage needs or with advanced multiport network requirements will find Mac Pro a more configurable and expandable system.

>>

[Steven H. Blackwell]

A MAc Mini is not s suitable FileMaker Server database server machine. It may function well as a file server, but not as a database server.

There will be more on this in a few days.

Steven

[dubl]

A client recently purchased a Mac Mini Server to run FMS v11.

Recognising that it is a 'pseudo server' and not an 'ideal' server setup, it is a cost-effective solution for the client. And yes, Mr. Blackwell, the cost of recovering data will more than outweigh the savings, but this is what the client now has.

I am trying to set up a backup schedule to the second drive inside the machine.

The FM KB article says:

ISSUE:

FileMaker Server cannot backup to a remote volume.

RESOLUTION:

Because of the increase in security changes that are done on the Operating System, this feature is no longer possible to maintain.

Regardless of the platform, we highly recommend that the files be backed up through FileMaker Server locally on a hard drive connected directly to the machine.

But this is NOT a remote volume - it is directly connected to the machine.

The volume name is 'betsserver', the second drive is 'BackupHD' and it has a 'Daily' folder in the root.

Paths I have tried which are NOT valid paths are:

filemac://betsserver.local/BackupHD/Daily/

filemac:/betsserver.local/BackupHD/Daily/

filemac:/betsserver/BackupHD/Daily/

filemac://betsserver/BackupHD/Daily/

filemac:/BackupHD/Daily/

filemac://BackupHD/Daily/

Am I wasting my time here?

Any other suggestions?

Thanks

Steve

[Cortical]

no, you are not wasting your time. You need to use the FM scheduled backup files, as the source for the system level backup, i.e. do not try to schedule the FMS backup to anywhere other than the default. I have not tried the following to an actual internal 2 HDD mac mini drive, but it works to usb|FW externals, self (shared...), LAN etc.

You can use ditto command to explore the path requirements, but recognise that the spaces in the path require escaping.

To play with the paths, a quick and easy way to capture a path is to drag the file or folder onto an open terminal window, you can then copy|paste the path

A timestamped zipped command would have a format like:

ditto -c -k -rsrc /Library/AAA BBB/CCC /Volumes/cortex/Users/Shared/fms_backups/`date +%y-%m-%d`_ccc.zip

or:

to -c -k -rsrc /Library/”AAA BBB”/CCC /Volumes/cortex/Users/Shared/fms_backups/`date +%y-%m-%d`_ccc.zip

A convenient way to achieve a backup is to write a shell script (e.g. bup_test.command), editing the details as required

SRC=/Library/AAA BBB/CCC/

DST=/Users/Shared/fms_backups/

TODAY=$(date +%y-%m-%d_%H%M)

ditto -c -k -rsrc "${src}" "${dst}${today}${now}_test.zip"

Variations on syntax have repercussions.

bup_test.command will be executable ( given appropriate permissions) by double clicking.

Also be aware of the 'permissions bubble' of the FM scheduled backup. You will (probably) need to write correct permissions to the ditto'ed output ; BatChmod is handy to explore this.

With a working shell script, writing a simple launchd (daemon) to schedule execution of the shell script, can be done with something like Lingon, which creates the appropriate xml from a straightforward interface

an example:

ml version="1.0" encoding="UTF-8"?>

KeepAlive

Label

com.cortical.bupHourlyLaunchd

ProgramArguments

/Users/admin/Library/Scripts/bup_fms.sh

RunAtLoad

StartInterval

1800

FMS scheduled backups are made into default /Library/FileMaker Server/Data/Backups.

Shell script saved into ~Library/Scripts/

Edited November 22, 2010 by Guest

[Cortical]

the fly in the ointment: the destination directory needs to have the correct permissions e.g.

fmsadmin read & write

[Justin Grigg]

A MAc Mini is not s suitable FileMaker Server database server machine. It may function well as a file server, but not as a database server.

There will be more on this in a few days.

Steven

I am at a location that is currently running a Mac Mini Server to serve up a 5 location business (spread out over 100 mile radius and running 8 databases) with 0 issues. We purchased it for a test run, but a year later and we couldn't be happier.

Just my 2 cents....

[John May - Point In Space]

Note that if outsourcing services is an option, we offer OS X based dedicated virtual servers starting at $99/month.

We are serious about the specs of our VM servers for optimum performance, unlike many other providers'

implementations. The hardware consists of:

- 8 core Xeon Xserves, at least 32GB RAM each

- Vtrak E-Class Fibre Channel SAN storage system (w/15K SAS drives)

- Redundantly pathed Fibre Channel network via multiple 4Gb links

- Gigabit switched ethernet networking

Our current path going forward is to use Xserves until they become impractical, then transition to Mac Pros, or

whatever other server-class hardware may become available between now and then.

Please contact me off-list at jmay(at)pointinspace.com for more specifics.

- John

[TheTominator]

A MAc Mini is not s suitable FileMaker Server database server machine. It may function well as a file server, but not as a database server.

There will be more on this in a few days.

Steven

More than a month has passed since your brief post.

Do you have any more information to add to support your assertion that a Mac Mini is not a suitable FileMaker Server?

-Tom

[IdealData]

Surely much depends on the demands you will place on the machine.

I have previous model Mini, 4Gb RAM and 64Gb SSD - runs a dream with over 30 users and 2.2Gb of data. The boot time is incredible - 8 seconds from COLD to desktop! (Although I don't have to restart it at all). Of course the mini is dedicated to FMS.

The SSD is key to all this as the original 5400rpm drives are gonna die pretty quickly if used as any form of server.

I run frequent backups and can crank up a Virtual Machine very quickly if anything fails.

This replaced a G5 Xserve.

[RodSierra]

A MAc Mini is not s suitable FileMaker Server database server machine. It may function well as a file server, but not as a database server.

There will be more on this in a few days.

Steven

Been quite some time since this was posted, had a prospective customer refer me to this post. Having about 6 Mac Mini's running as FM Servers with different clients I'd like to get a response to this post for future reference.

[Wim Decorte]

Been quite some time since this was posted, had a prospective customer refer me to this post. Having about 6 Mac Mini's running as FM Servers with different clients I'd like to get a response to this post for future reference.

Cheap hardware is cheap for a good reason: it's quality is inferior to server-class hardware. Inferior quality of the hardware translates in shorter mean-time-between-failures or in other words the Mac Mini is much more likely to fail a lot sooner than server-class hardware. When it does fail and you don't have a good backup strategy your clients are at risk of losing everything or at the least incur long outages. Every deployment is specific so the questions that needs to be asked are:

- how much does an outage cost the company?

- how much does it cost the company to lose everything?

The increased risk of using a Mac Mini comes at a cost. Is the cost differential between a Mac Mini and a real server worth the increased risk?

Note that I said "more likely to fail". Not "will fail". You'll hear stories of people running the Mac Mini for years without issues. They're just lucky but the odds are not on their side.

Like Steven, a substantial part of my business is in rectifying FileMaker Server deployments that have gone bad. I have yet to see one deployment where the increased risk was worth the Mac Mini "cost savings".

[IdealData]

Been quite some time since this was posted, had a prospective customer refer me to this post. Having about 6 Mac Mini's running as FM Servers with different clients I'd like to get a response to this post for future reference.

The hardware deployment has to be put into perspective to the installation and budget constraints. FM Server is deemed necessary where there are more than 5 users and the same will handle 100s of users. So there's some scope!

A 5- 10 user system is unlikely to have the same budget allowance as a 100+ user system.

The likelihood of hardware failure is not simply governed by manufacturer tolerances and MTBF figures - there's also the workload it is subject to, the conditions it is housed in (air con rack?), and the stability of the electricity supply. A UPS will help greatly in areas subject to power outage, and need not be expensive to keep a single server alive for short durations.

The hard drive in the Mac Mini is not suitable - replace with an SSD. 4GB of RAM also.

The cheapest security measure is a good backup strategy - and test it!

Interestingly the XServe Transition Guide suggests the OS X Server Mac Mini is fine for a 50 user workgroup, but can also be deployed as a single-task server to larger numbers.

With current pricing you can get nearly 4 Mac Minis for 1 Mac Pro. At that price you can afford a completely redundant Mac Mini, waiting for the other one to fail! Should be up and running in the same time it takes to recover the last backup.

[Cortical]

A Mac mini is a very useful box to use as a FileMaker server, although it meets the FM requirement it does have some limitations though.

One needs to be aware of the limitations.

1. Hard drive(s)

There is a difference between server grade HDD and consumer grade HDD, both in MTTF and speeds. HDD will fail, that is a fact of electro-mechanical life.

A strategy is therefore required to 'avoid' mechanical failure (e.g. SSD, or to deal with it when it happens (replicate system)...

Splitting out the HDD to external enclosures and employing server grade HDD while just using the mm mobo is limited by connection speed, and probably somewhat fanciful.

2. RAM

servers would typically utilise ECC RAM

3. PSU

servers commonly have dual power supplies. The Mac mini has one external PSU.

4. HEAT

Mac mini's are fanless, they get warm, place them on the wrong sort of surface (e.g. a wooden shelf) and they get even warmer; raise them off the surface 10-20 mm will make a difference.

Place an array of them in a 1U and a variation between mini's is probable.

In a cooled server room, heat dissipation does not seem to be a problem with a standalone Mac mini.

Others may have other observations to add to this short list.

Mac Mini's can function well in high load high access environments (think Hospital Emergency Dept. PMS), but it is pushing the performance envelope. Pushing does not mean breaking.

Provided one is prepared/geared to handle a failure when it occurs, a. Mac mini is a very practical choice. I would rather be running Unix (Mac OSX) than Windows, and a Mac mini than equivalently priced pc hardware any day.

I am sure we would all be running state of the art top of the line maxed out hardware (of whatever platform persuasion) if we could. But we can't, so we don't.

It is horses for courses. And Pragmatism generally wins.

Given that EVERY system can have catastrophic failures, an appropriate DOCUMENTED and TESTED backup/recovery system/protocol is MANDATORY.

I was talking to a senior systems guy only yesterday, a major hotel chain high end clustered system running Oracle, fell over. So did the fallback/backup system. 2 weeks of hell.

Can't happen to me? Oh yes it can. The hardware is only one link in the chain.

[Cortical]

new mutterings:

What replaces the Xserve?

http://www.9to5mac.com/40192/what-replaces-xserves

<<

Apple caused quite a stir when it announced the discontinuation of its Xserve line on November 5th. Mac IT departments immediately went scrambling for new solutions. Apple officially offered up a faster Mac Mini as a low end replacement and the Mac Pro as a high end replacement, both with Snow Leopard Server installed. IT professionals who maintain data centers small and large know that these aren’t complete solutions compared to the density of the Xserves.

So, what is everyone doing? CNET quotes the Enterprise Desktop Alliance that says about a third of Xserve shops will be migrating to other hardware within the year. That reaction might be a little extreme for Apple because we’ve heard that Apple has been sending out reps to quell the exodus.

One Apple Enterprise shop we spoke to said that Apple’s System Engineers have been on a campaign to assure that Apple has another solution in the pipeline beyond what they’ve announced. There wasn’t an exact timeline of release but there was an indication that when the Xserve inventory runs out, the new product will be released. That means a new Enterprise product could be ready in February.

We’ve got nothing beyond that, but there are a lot of new variables in the ecosystem. Apple builds its own efficient A4 processors now which, running in parallel, could match the current Xserve performance without using nearly as much power. Apple is also building a data center in which it could offer enterprise cloud services.

Don’t forget those Virtual Machine clues we found last month.

Finally, Lion is due later in 2011 but it isn’t yet guaranteed that Apple will continue to build Server versions of the Mac OS.

>>

[RodSierra]

All very good info, thanks for filling in the blanks. The installations we have using mini's are all less than 10 users, and all have tried and tested external back up strategies that we monitor for the clients. We also specify only server grade hardware for any of our installs that require it based upon user count and activity. We are very happy with both our high end server installs and our mini's in small solutions. But as has been stressed earlier by many, a robust backup strategy must be implemented, and monitored in all cases. Economics plays a big part in the selection of the solution, but cannot be the only factor. Does Apple publish MTBF data on any of their current products? I've not looked that hard myself for it, but it would be interesting to compare.

[Wim Decorte]

Provided one is prepared/geared to handle a failure when it occurs,...

And this is the key. A deployment is not just a server and software. It's a whole disaster recovery plan that should be evaluated constantly and tested at least once year.

Sub-standard hardware increases the risk of failures so this need is even more crucial there (and reduces the perceived cost savings).

As I stated in my opening remark at last year's devcon: "we're good developers, but are we good deployers?" Sadly the answer is frequently no. Even though an unstable deployment WILL look very bad on us.

For those developers who can't/won't/choose not think about all the deployment implications there are the best practices.

I'm not saying that we should all use top-notch super-redundant servers. But we should look a bit harder at the often false immediate cost savings of using cheap hardware.

[El_Pablo]

Just to let you know that we are running FMS11 on a Mac Mini server since July 2010 and no problem here. We are 10 users and I could handle up to 25 users. The server is dedicated to FMS only.

[jfletch]

This is so sad. Guess we all are gonna have to make peace with Windows server hardware. Much as it pains me to say it. I used to paraphrase the old bumper sticker: "I will give up my Mac when they un wrap my cold dead fingers from around it." Now, while I still can use it for client machines, servers are less and less an option. I worry about what the FileMaker hosting companies are going to do.

Hackintoshes anyone?

[dsghs]

A MAc Mini is not s suitable FileMaker Server database server machine. It may function well as a file server, but not as a database server.

There will be more on this in a few days.

Steven

I'd like to know if this is still your opinion, Steven. I've hosted a solution with two mac minis (main/client and web publishing) for over a year now with the only problem being a third-party SSD failure due to heat problems. A crucial engineer also told me that an SSD wouldn't be the best choice for a server hd due to garbage collection issues. I have since setup RAID mirroring - both mini servers with trad 7200rpm drives and the little cluster performs beautifully.

Why is this setup not practical?

[Wim Decorte]

I'd like to know if this is still your opinion, Steven. I've hosted a solution with two mac minis (main/client and web publishing) for over a year now with the only problem being a third-party SSD failure due to heat problems. A crucial engineer also told me that an SSD wouldn't be the best choice for a server hd due to garbage collection issues. I have since setup RAID mirroring - both mini servers with trad 7200rpm drives and the little cluster performs beautifully.

Why is this setup not practical?

It's not a question of "practical", it's a question of risk management. If the solution is business critical then you want to minimize downtime and minimize the chances of data loss. Using an inferior machine increases the chances of both of those happening. Meaning that you have to spend more money elsewhere to mitigate that risk. At that point it becomes a question "can we save that expense by getting a server that is more stable?"

Note again that this is about risk. I've stated earlier in this thread that using a Mac Mini is not going to crash because it is a mac mini. Just that the quality of the hardware makes it more likely.

More often than not, the use of a Mac Mini is an indication of larger "oversight": not spending enough time thinking about disaster recover strategies. And when the machine conks out it frequently exposes the lack of DR thinking: no backups or invalid backups, scrambling to find a replace machine, no documentation on the setup... all leading to expensive downtime and potentially expensive data loss.

Business critical solutions should be treated as such.

[dsghs]

It's not a question of "practical", it's a question of risk management. If the solution is business critical then you want to minimize downtime and minimize the chances of data loss. Using an inferior machine increases the chances of both of those happening. Meaning that you have to spend more money elsewhere to mitigate that risk. At that point it becomes a question "can we save that expense by getting a server that is more stable?"

Note again that this is about risk. I've stated earlier in this thread that using a Mac Mini is not going to crash because it is a mac mini. Just that the quality of the hardware makes it more likely.

More often than not, the use of a Mac Mini is an indication of larger "oversight": not spending enough time thinking about disaster recover strategies. And when the machine conks out it frequently exposes the lack of DR thinking: no backups or invalid backups, scrambling to find a replace machine, no documentation on the setup... all leading to expensive downtime and potentially expensive data loss.

Business critical solutions should be treated as such.

Perhaps when more users come about I'd consider the $3k machines. The difference between consumer and "server-grade" equipment really has become smaller though. For instance, the E55xx chips are identical to the i7 macmini stock. The only difference is a second QuickPath interconnect and slightly different instructions. Reliability is the same. It's way different than only a few years ago- we've come a long way and the differences are becoming negligible.

[Wim Decorte]

we've come a long way and the differences are becoming negligible.

I highly doubt it. You can't go by chips only because that's just another "performance-only" bias. MTBR of all the other components are just as important.

I'm not here to convince you or to preach. Just sharing my experience.

Actually I find these kinds of discussions very weird because it seems so obvious to me - having worked through so many disasters - that server class hardware is just better than consumer class hardware. The point hat I've been making at the last 5 or so developer conferences is "we are good developers, but are we good deployers?" Most of us seem to be afraid to add to the proper deployment cost to the cost of the solution for fear that the client will run away. My experience is that they won't, provided that you can explain why a solid deployment is just a crucial as solid development practices.

[Steven H. Blackwell]

I want to echo what Wim says here about using hardware adequate to the task. When the head server engineer of FMI stands up on the stage and cautions about the use of consumer grade equipment, I tend to listen. Most of these consumer grade machines are tested for 8 to 5, 5 day a week usage.

Wim's point about becoming good deployers is critical. It's only going to get more complex in the future, so I strongly recommend starting on that path to good deployment now.

Steven

[Joost Miltenburg]

Just installed one of the new late 2012 MacMini servers ( standard config ). From a performance point of view I am very pleased ( and this is with the standard hard disks and just 4 GB Ram. ( The Amsterdam Apple Store didn't have the memory ons hand . The i7 is humming along nicely. The second hd is a large backup disk.

 

Maybe we will go with more memory and a ssd later on. Had a MacMini with a failing HD once, but that can happen, right ?