Test for a new NAS - Recommendations

[WinXaito]

Hello,

I set up my NAS last week. I am now in the testing period.
I have already done some memtest (with memtest86+), about 24 hours of testing without error.

For the disks, I saw that there was the solnet-array-test from @jgreco.

What do you recommend for the other tests?

- For CPU Stress?
- For disks?
- Other things?

And how long should the tests be run (I did see the document that says 1000h of recommended tests, but as a private person I find that way too much...)

Thanks

 

[jgreco]

As a private person, perhaps you have access to equipment that isn't subject to the normal laws of physics and statistics. Cool! ;-)

1000h covers important issues such as HDD infant mortality, CPU heatsink compound curing/break-in periods, and several other issues.

memtest86 on large memory systems often don't even complete a single pass in a measly 24h.

Arctic Silver 5, for example, has about a 200h cure time, which is helped along if you can exercise the CPU during that time.

 

[WinXaito]

Thank you for this information.

The first pass of memtest takes about 3 hours, the following about 8 hours. So I'll do a few more passes.

Concerning the CPU, I installed a Windows and did load tests with OCCT. It seems to hold, maximum temperature obtained so far: 82°, average temperature 77.50°.
It seems OK.

About the thermal paste, I left the one that was on the cooler included with the processor (from Intel)

-----------------

Regarding your tool (solnet-array), must the disks be in a pool? Or must they not be? Or whatever?

If I'm not mistaken, it's all about reading, so nothing is affected at the disk level?

And last question, if I had to interrupt the test (for some reason), could it be dangerous for the disks? Or just annoying because I will have to start again?

 

[jgreco]

it's all about reading, so nothing is affected at the disk level?

Yes, it's safe to use on any set of disks, regardless of whether they're in a pool or not. The tool predates ZFS and was designed for parallel SCSI bus/controller testing, and disk burn-in as a side effect. It doesn't know or care about ZFS. It will, however, significantly impact performance of ZFS as it runs lots of IOPS at the drives.

It is written entirely in shell script, and I think the worst that could happen would be that some dd processes could be left running. These are what do the actual disk input, and the script is just managing the overall orchestration and statistics reporting. Since it is just doing reads, I cannot envision any negative outcome to your disks other than being slow while dd's finish running.

 

[WinXaito]

One more question, I see on truenas scale that APT was blocked (and of course, I understand why).

But your script need *camcontrol*. So, do I have to unblock apt and install camcontrol or run your script in a virtual machine ?
I think the VM is preferable, but I don't know if a VM can access to the host disks and if the performance is okay.

Otherwise I will unlock apt, install camcontrol and after that maybe reinstall truenas...

 

[Etorix]

The script is for FreeBSD, hence 'camcontrol'.

 

[WinXaito]

Oh yeah I see.

So, there is an other script for linux-line OS ?
Otherwise I can edit the script for match with debian OS. Or maybe install à truenas core for run test :) I will check.