suggestions for a low power system to replace that old dual xeon macpro

[hotdog]

UPDATE:
I finally found time to measure power consumption on the mac pro:
it's idling at 205 W.

I ordered a X10SDV-2C-7TP4F, 2 sata doms and 64gb of ram as a start.

 

[John Doe]

thanks for the update, please continue updating us on your build.

Would like to see the final measurement (power) and some performance pictures

 

[Jessep]

it's idling at 205 W

That's likely an idle power usage of 2-4x a more modern system.

 

[hotdog]

thanks for the update, please continue updating us on your build.

Would like to see the final measurement (power) and some performance pictures

I will.

That's likely an idle power usage of 2-4x a more modern system.

That was my inital guess too.

I am tending to go with 5 x Crucial MX500 2TB in a raidz1 config.
the data pool will be backed up and we could sustain a total loss, since ongoing projects are always on at least one workstation.

Do you agree such a pool will have a better performance than the striped tripple mirror vdevs made out of 6x 2.5 4TB Barracudas?

 

[Jessep]

Crucial MX500 have a known issue with FreeNAS.

 

[hotdog]

Crucial MX500 have a known issue with FreeNAS.

hmm. tnx for pointing this out.

reading this thread...

Is my available space correct?

Good day, I have a feeling I might be missing around 1TB of usable space on my CIFS Share. I am running Freenas 9.10. I got 4 x 4TB drives running in RaidZ1. when looking at Storage -> Volume , it looks like this: The top volume has 3.4TB available, but the second volume only has 2.2TB...

www.ixsystems.com

... I must confess that I have forgotten about that 50% rule (for mirrors) and other general considerations like block size.

With 47% usage our current storage pool (of 4TB) is scratching on that limit, but doing no snapshots and basically adding data, overwriting* only a small amount of working files at the moment I don't think we would be affected by any performance loss going over 70% but that's a wild guess only.

* I learned zfs will not overwrite but choose a new location for the data block then because of its copy-on-write behavior.

typical folder is looking like that, it's really mixed, doing 3d stuff, single files can to be larger (up to 2Gb):

As I understand MIRRORs are better in handling such data?
AND/BUT:
The plan to use snapshots will fragment the drive because of the releasing will force me to stick to that 50% rule for mirrors, right?

so: maybe I better stick with the spinning drives and get more raw space for the money.

 

[volothamp]

Crucial MX500 have a known issue with FreeNAS.

can you be more specific? I’m thinking of a crucial Mx 500 build

 

[Greg_E]

Other than ZFS doing full write before erase and using up all the extra blocks assigned for repair, I don't find anything specific to MX500 with a quick google. So I'd like more specifics too. The closest reference I found was a user saying that they've had them going for around 3 years and that SMART reports that no repair blocks are left, which means the next time a test is run and there are bad blocks, you lose capacity going all the way to complete failure. I have some MX500 in desktops, old units like 250gb that still work, but their usage is much different from a NAS, and especially ZFS.

 

[volothamp]

Other than ZFS doing full write before erase and using up all the extra blocks assigned for repair, I don't find anything specific to MX500 with a quick google. So I'd like more specifics too. The closest reference I found was a user saying that they've had them going for around 3 years and that SMART reports that no repair blocks are left, which means the next time a test is run and there are bad blocks, you lose capacity going all the way to complete failure. I have some MX500 in desktops, old units like 250gb that still work, but their usage is much different from a NAS, and especially ZFS.

There's also this

What does critical error "Currently unreadable (pending) sectors" mean?

So FreeNAS threw this at me this morning: "CRITICAL: June 9, 2018, 9:57 a.m. - Device: /dev/ada3, 1 Currently unreadable (pending) sectors" smartctl -t long returned: smartctl 6.6 2017-11-05 r4594 [FreeBSD 11.1-STABLE amd64] (local build) Copyright (C) 2002-17, Bruce Allen, Christian Franke...

www.ixsystems.com

the problem with a specific SMART error to be ignored but to me it's just a small detail.

Greg tell me how your build goes, I'll do the same in a few months.

 

[Constantin]

I would also consider how the coming special VDEVs (TrueNAS 12.0) might influence your system design and pool layout.

Alternatively, You could consider a small pool that consists of SSDs (mirrored) to allow fast reads and writes with 10+GbE infrastructure while maintaining a bigger spinning pool for long term storage.

 

[hotdog]

UPDATE:

The build is complete and up for about 3 month or so.
I did a complete burn-in as described somewhere here in the forum.
Fan script is running to take care of the HDDs.
Everything works great.
The Server takes 52W while idling.
Feels snappy.
I haven't upgraded to 10GB yet, so bottleneck should be the network infrastructure at the moment.

Here again the parts:

mb - SUPERMICRO X10SDV
memory - Samsung 32GB M393A4K40BB0-CPB
case - Micro-100B Mini Server Gehäuse
psu - Seasonic SSP-300SUG
hdd bay - SilverStone SST-FS212B
pool hdds - 6x SEAGATE Guardian BarraCuda 4 TB
freenas on 2x 64GB Supermicro SATA Doms
jail ssd - SSD WD Green PC 3D, 240GB

a 12cm Fan blowing down on the board
a couple of Noctuas
thin Noctua NF-A12x14 as air intake for the hdd bay (some modding involved)


tnx again for all your suggestions and kind help.

ps: I did return the fractal node, I just didn't like the look of it.

 

[kiriak]

when you say 52 W idling,
you mean with the HDDs spinning or what?

 

[hotdog]

when you say 52 W idling,
you mean with the HDDs spinning or what?

yeah. remember these are 2.5 inch hdds. according to the datasheet they consume read/write 1.9/2.1 W and just spinning 1.1 W each.

 

[volothamp]

How's the noise @hotdog?

 

[hotdog]

How's the noise @hotdog?

the server is very quiet. we do not have a seperate room for it so this was a must. The fan coming with the silverstone hdd bay was very loud. I replaced it with noctua but found out that the cooling of the drives wouldn't be good enough for the hotter summer days. That's why I decided to mount that intake fan directly in front of the bay. Together with the fan script the temps now are very good at very low noise level.

 

[hotdog]

Update:
my jail ssd (SSD WD Green PC 3D, 240GB) failed a couple of days ago.
I am not sure why - but I had my system dataset on it. (kind of stupid).

But I guess I was lucky, I was able to move the location of the dataset to the main data pool on the spinning hdds. The logging history is gone, but the system seems to work normal again.
the fan script with its logging file also was on the that ssd, I also moved that one.

Now the questions:
system data set: better put on the mirrored 64GB sata doms?

I am not running other jails or stuff at the moment but maybe want to use the nas as a vpn server, because the fritzbox vpn is too slow. So a replacement of the failed ssd is an option, too.
Would you guys then go for a mirrored solution to also put the system dataset there again?

 

[Constantin]

SATADOMs are expensive but they work very well for system datasets. I use a mirrored boot pool even though that may not always save your bacon due to the limitations imposed by the BIOS. If disk 1 in the boot order always fails somewhere partially into boot, then the NAS will be done until you pull it. Recovery from failure with a mirrored boot pool is quicker though, which is why I use it.

 

[hotdog]

SATADOMs are expensive but they work very well for system datasets. I use a mirrored boot pool even though that may not always save your bacon due to the limitations imposed by the BIOS. If disk 1 in the boot order always fails somewhere partially into boot, then the NAS will be done until you pull it. Recovery from failure with a mirrored boot pool is quicker though, which is why I use it.

tnx for the answer Constantin.
So just to be clear, you suggest to just move the dataset to the sata doms where the freenas system is located then?

 

[Constantin]

I'm not sure I understand your question or I am misinterpreting, and neither is a good thing.

I would use the SATADOMs as intended - to hold the various system images (i.e. "TrueNAS 12.0 U1") and the system swap files. Nothing more. SATADOMs are not particularly fast SSDs, nor do they tend to be big. I'd dedicate them to the system files needed to boot and run a NAS.

 

[hotdog]

I'm not sure I understand your question or I am misinterpreting, and neither is a good thing.

I would use the SATADOMs as intended - to hold the various system images (i.e. "TrueNAS 12.0 U1") and the system swap files. Nothing more. SATADOMs are not particularly fast SSDs, nor do they tend to be big. I'd dedicate them to the system files needed to boot and run a NAS.

ok maybe to be more clear:
I am not really sure how to configure this setting (System Dataset Pool):

7. System — FreeNAS®11.3-U5 User Guide Table of Contents

www.ixsystems.com

now it's pointing to the big pool (and before to the single ssd that failed), not the satadoms.
so I guess this is ok then.
I thought 64GB sata doms should be big enough to also hold the system dataset.