FreeNas on leftover hardware

[evryc]

Hi,

since my offsite backup server is lacking sufficient space and is running windows (Athlon II X4 and 8G DDR3) at the moment I plan to "upgrade" it a bit. Unfortunately I don't have any budget so I thought about building something with leftover hardware.

Since this is just my backup server he doesn't run 24/7, but will be automatically started once a week and shut down once the backup job is finished. Because it would be pointless to have him standing near my production servers in the basement I have him running about 50 meters away connected via a single gbit cable.

I planned on using this hardware, everything besides the usb thumbs is already in my Possession:

2 * 16gb Sandisk USB
7 * WD RE4 2TB WD2003FYYS
1 * Supermicro X7DVL-E with IPMI card
2 * Xeon E5405
6 * 4GB DDR2 ECC Kingston
1 * Silicon Image SATA PCI Controller with Passthrough

I'd get about 12TB RaidZ Volume, 24GB ECC DDR2, 8 Cores.

I am bit sceptical about the DDR2 RAM so i'd be glad to hear an experienced opinion on that matter, keeping in mind that the maximum transfer is capped by an single gbit interface anyway.

With kind regards
evryc

 

[Ericloewe]

I am bit sceptical about the DDR2 RAM so i'd be glad to hear an experienced opinion on that matter, keeping in mind that the maximum transfer is capped by an single gbit interface anyway.

It won't hurt to try, but don't be surprised if the server is sluggish. The FSB on that thing is a big bottleneck for ZFS.

 

[BigDave]

Since this is just my backup server he doesn't run 24/7, but will be automatically started once a week and shut down once the backup job is finished.

The practice of turning off the server in between backups, kinda defeats the whole
purpose of pool scrubs and SMART testing.
So be forewarned, you may one day hit the power button and find out that bit rot
has rendered your backup unavailable and unrecoverable.
Yoda says, "Sad for you, we will be then!"

 

[evryc]

Hmm, should I use another Filesystem or add some maintenance operations to the backup job? At the moment I am backing up tar and winrar archives with parity, because of NTFS. Should I continue this method even with ZFS? I hoped a bit that I would get rid of archive necessity with ZFS.

 

[BigDave]

Hmm, should I use another Filesystem or add some maintenance operations to the backup job?

I think you might benefit from reading this ZFS Primer
and yes, the maint. operations are needed for data integrity.

 

[evryc]

I have read him, but I didn't realized the self-healing operations weren't executed on runtime.

 

[BigDave]

I have read him, but I didn't realized the self-healing operations weren't executed on runtime.

SMART testing is highly recommended, but has to be set up on a schedule.
A scrub is defaulted to occur every 35 days (I run mine every two weeks), but that is recommended
to occur during off peak hours as it's a heavy read operation (if it finds no errors to correct).
I guess you could run a scrub manually every two weeks (after ever other backup), I don't have any idea
what you would do with smart testing though. If your machine runs 24/7 all this stuff is pretty easy,
but shutting down the server makes for a big PIA hassle if you do things the way you should.

 

[jgreco]

I have read him, but I didn't realized the self-healing operations weren't executed on runtime.

Yes, self-healing operations are always performed as data is accessed, but there's still a lot of value in running things like scrubs and monitoring SMART, because each of these things protect against different risks. You want your car to have seat belts AND air bags AND crumple zones. Each of those things saves your life in a different way, and not all of them come into play for every accident.