Pulled a disk for testing, web UI not responding.

[diskdiddler]

NOTE: Data not important on the server, just testing purposes.

Using a HP Microserver N40L with 4gb of ram, array was 3x2TB disks - it was mid copy.
The copy is still working in teracopy to my Windows PC right now, the data is clearly fine - but the webUI is not responding. Is this normal? Does the CPU freak out for half an hour or something when a disk drops out?

Curious how to interpret this, machine is otherwise stable, CPU isn't a beast in the Microservers but it's not too bad. AMD Turion 1.8ghz I think

 

[diskdiddler]

Placing the disk back in the server has caused the file copy to fall over entirely.
Web UI is still dead
Machine responds to ping.

 

[BigDave]

Please read this before posting
The community needs more details to assist you.
Welcome!

 

[diskdiddler]

Sorry I forgot to mention latest version, running from a USB stick, I believe it's 9.2.1.6? Drives are mixed 2TB SATA models.
I do not know which configuration I'd set them up with, I chose the optimal layout for 3 disks when configuring for testing.

 

[cyberjock]

Well, first off you need to upgrade your system to meet the minimum requirements. Otherwise any problems may or may not be because of not having enough RAM.

 

[diskdiddler]

I've seen posts all across the internet about it running with 2gb and 4gb.
The system will eventually have 8gb of ram. Regardless I wouldn't expect the webUI to stop functioning, I'd expect disk performance to get far slower instead, so I can still administer the system.

 

[cyberjock]

Yes, and those people making those claims are referring to the old FreeNAS .7 project which has no codebase related to FreeNAS 8+ or they are idiots. Our wiki pages make it clear that you need 8GB of RAM else you risk pissing off the server gods who will smite you with random unexplainable errors until you slit your wrists, give yourself up sexually to the server gods, or go with a different OS.

 

[diskdiddler]

If what you say so vehemently is completely factual, the software should literally not allow users to configure with 4gb, period.
It simply shouldn't boot if it's normal behaviour for the OS's web UI to fall over due to a disk removal.

 

[BigDave]

He said "smite you with random unexplainable errors"
Sounds to me like that's what is happening to you now ;)

 

[cyberjock]

No, you aren't understanding. There are 2 potential problems at work:

1. You don't have enough RAM to expect reliable operation.
2. Your hardware (and/or drivers on FreeBSD) may not support hot-swap and/or hot-plugging.

If #1 is actually your problem (there's no way to prove either way except to upgrade to 8GB of RAM) then you are going to have random problems until you upgrade. If you are using UFS then you don't need 8GB of RAM and 4GB will *usually* work. 9.3 will have no UFS support, will be x64 only, and will boot up with a warning if you don't have the minimum amount of RAM.

If #2 is actually your problem (and your low RAM is not a problem at all for your exact scenario) then the solution is to get better hardware. I will tell you that the N40L is widely used around here and since I don't own one I can't really vouch for if it handles hotswap or not. I also cant vouch for if there's some necessary BIOS setting you didn't set, some BIOS version needed, etc.

It is *very* well known that *many* systems people build, especially those that aren't server-grade, don't do hotswap and/or hotplug outside of Windows. Because of this very well known and very well documented problem I never recommend people do disk swapping with the system on and to *never* do it under *any* circumstances until you've tested your system and validated that it is safe to do. It sounds to me like you've just tried it and it didn't go well, so you are left with 2 options:

1. Figure out if something is actually wrong and it *should* work.
2. Accept the limitations you have.

So, in order of operations of what I'd do in your shoes:

1. Upgrade to at least 8GB of RAM (16GB would be better and is supported on N40L despite what the manual says).
2. Check your hardware out and see if it actually *should* work with FreeBSD. Not all hardware works properly in FreeBSD.
3. Accept your limitations and just don't pull drives with the system on.

 

[diskdiddler]

Well the final machine is built now, with 6x5TB Toshibas, 8gb of ram and 9.2.1.6 - any tips on how to fully test the array before using it?

 

[BigDave]

Try this post

 

[diskdiddler]

Thank you.
Do you know why I might not be able to change the acoustic level? It's set to disabled at the moment, when I chose minimum it won't change. I've googled - can't find answers. The noise from all 6 disks writing is horrendous

 

[BigDave]

Thank you.
Do you know why I might not be able to change the acoustic level? It's set to disabled at the moment, when I chose minimum it won't change. I've googled - can't find answers. The noise from all 6 disks writing is horrendous

What your describing maybe suggests a failing unit or possibly more than one???

 

[diskdiddler]

It seemed rebooting fixed it, I did find a guide which said old versions required a reboot to change those settings, so I changed them, it still said "disabled" then rebooted, but after that changing the settings is now reflected in the disk menu.

Mind you the disks are still quite loud, I'd say it's simply because 6 actuators are moving at exactly the same time, "minimum" accoustic level did basically nothing :/ Bloody 7200rpm Toshiba, noisy junk.

 

[diskdiddler]

Also for anyone else curious, at least on these disks, the sound of "minimum" and "maximum" is completely identical.
EDIT
I can confirm this model of disk does not support AAM (acoustic management) and I would recommend people avoid purchasing them at all costs, it's 2014 and no sound adjustment? .... mind boggling. Also the website for support for them is TERRIBLY sparse for information / downloads etc.
Model#: Toshiba 5TB Internal MD04ACA500 - AVOID

 

[joeschmuck]

Changing the acoustic level slows down the head actuator in the drives which leads to less noise in general. This should not impact data access by an appreciable amount. What I find interesting is you needed to reboot to make the settings take affect because FreeNAS software was to be changed to make the hard drive settings an immediate setting, no reboot required. I'm curious if only the sleep timer was changed and the other settings were not.

When you do the burn-in testing, run Memtest86 for at least 24 hours. I use to say to run it for 3 passes but that may not be enough is the consensus. Also run a CPU stress test for no more than 2 hours.

I assume you will do some benchmark testing or at least be concerned with how fast your network traffic is flowing. Make sure you take into account all your hardware. The last throughput issue I helped someone troubleshoot turned out to be his main computer's internal laptop hard drive couldn't support the throughput, the FreeNAS system was running fine the entire time.

 

[diskdiddler]

I know what acoustic levels do (thank you though) however the disk literally doesn't support acoustic management, it's utterly completely mind boggling.
Considering any kind of RAID array will have multiple disks writing at the same time, it's to be expected actuator noise will be louder for any storage but the noises I'm hearing are just, not ideal. I can never heard my WD greens, they are abnormally quiet and it's ...well it's bloody great.

 

[cyberjock]

Well, those 5TB drives are monsters.. you don't buy them because they are quiet. You buy them because they hold more pron than any other drive out there. ;)

 

[diskdiddler]

Not at all, storage should be quiet, cool, reliable if it's on a platter. I don't buy for speed, I buy for reliability and noise.
My last 20 disks have been WD Greens and as I said, they are mind bogglingly quiet.

I've done what I can to hassle Toshiba, I doubt there will be any traction. Just an appalling product to not support AAM and a poor decision from the storage industry in general to continue to manufacture 7200rpm disks, people want space, not speed from their platters, sigh.