How to cold backup raw data without RAID.

[Wipi]

What is the recommended method to backup data to an offline build without regular update?

I thought about building a system with 1 HDD only No-RAID, taking advantage of ZFS or BTRFS metadata and data integrity checksum to avoid bitrot or other factors that may affect the written bits. Can FreeNAS do a scrub and fix errors using only metadata and checksum?

If it will not work, I may just do a RAID1 4TB+4TB using WD BLUE drives, and when a drive hangs due to TLER I'll just restart it manually. Maybe I have it all wrong... what do you think?

 

[Arwen]

My FreeNAS has 4 x 4TB disks in a RAID-Z2. My backup solution involves attaching a single
750GB or 8TB disk, formated with ZFS. (The 750GB does not backup my media...) Scrubs let
me know my data on the backup drives is still good. So, yes, their is a benefit to using single
disk ZFS pools.

Please note that data faults on single disk ZFS pool would cause data loss of the affected files.
Since my disks are used for backups, not a problem because the original data should still be
available somewhere, (original source or other backup disk).

Also note that ALL ZFS metadata has 2 copies, even on a single disk pool. Critical metadata is
supposed to have 3 copies. You can add some redundancy to the regular data in a single disk
pool by using ZFS' "copies=2". But, it's less redundancy since if the whole disk fails, everything
is gone.

Last, ZFS seems to be less affected by TLER issues than hardware RAID solutions. Plus, you can
do some tricks for head parking on consumer drives, (WD widdle3.exe program).

Edited: Corrected typos.

 

[Wipi]

Last, ZFS seems to be less affected by TLER issues than hardware RAID solutions. Plus, you can do some tricks for head parking on consumer drives, (WE widdle program).

Interesting info. In the past days, I learned about re-silvering and the stress HDDs go through to rebuild the new disk... I don't want to experience it. this made me change my mind about RAIDZ. I prefer to pay extra and go for RAID1.

I decided to 2x2TB in RAID1 for important data and 1x4TB for non important data. Thank you Arwin for pointing me towards WDIDLE3.EXE. I've seen it before but I wasn't sure it enables TLER. Did you actually test WD BLUE hack to enable TLER?

Some BTRFS user was against trusting regular on board ports, instead use an HBA to minimize the layers. What do you think about getting a HBA?

 

[Arwen]

...
Thank you Arwin for pointing me towards WDIDLE3.EXE. I've seen it before but I wasn't sure it enables TLER. Did you actually test WD BLUE hack to enable TLER?

The WIDDLE3.EXE program only reduces auto-parking of the heads during idle time. It does nothing
with TLER as far as I know. Constant parking of heads is not a good thing for a NAS that is always
powered on. You can get 100,000 parks in just 6 months.

TLER is Time Limited Error Recovery, (some vendors use different names). Basically, on reading a bad
block, consumer drives go through a long process to try and recover the data. This can be bad for
hardware RAID contollers, because they think the drive has died. ZFS is not affected by this as much.

Some BTRFS user was against trusting regular on board ports, instead use an HBA to minimize the layers. What do you think about getting a HBA?

Regular system board SATA ports should be fine. But, FreeBSD, (used as the kernel with FreeNAS), does
not have great support for lesser brands and models of SATA chips. Thus, we recommend that you to do
your research on any boards and their SATA, (or SAS), controllers. For example, Intel SATA controllers
work quite well, as do LSI SAS controllers.

Edit: Corrected typos.

 

[wblock]

In the past days, I learned about re-silvering and the stress HDDs go through to rebuild the new disk... I don't want to experience it. this made me change my mind about RAIDZ. I prefer to pay extra and go for RAID1.

There is no difference between these as far as resilvering is concerned. If a disk fails, enough data must be copied to the replacement to make it a copy of the original. ZFS actually copies less data than traditional RAID because it only copies allocated blocks.

Motherboard ports, at least with Intel chipsets, would be my first choice. After that, LSI 9211 HBAs are fine, but don't add a board that can convert an addtional 30 watts of power into heat unless you need it.

This is the second time this week that the idea of motherboard SATA ports being somehow lesser has come up. Is that a Windows or Linux thing, or maybe from off-brand chipsets?

 

[gpsguy]

Probably the off-brand chipsets, like Marvell.

This is the second time this week that the idea of motherboard SATA ports being somehow lesser has come up. Is that a Windows or Linux thing, or maybe from off-brand chipsets?

 

[Wipi]

ZFS seems to be less affected by TLER issues than hardware RAID solutions.

Does this mean I can get away with buying the cheaper WD Blue?

 

[Robert Trevellyan]

Depends on your requirements. Sometimes a failing desktop drive will make the whole system unresponsive, which makes diagnosing and fixing it much harder.

 

[Stux]

For a backup drive, I'd consider ZFS so you can integrity check it, and then rsync your files to it, perhaps even with snapshots.

The backup is for if your primary NAS has a serious failure.

If you're only connecting the backup drive irregularly then I'd think a blue is fine, but ensure you scrub the disk as part of your backup regimen

And I'd organize some sort of online offsite backup task for your most critical data.

 

[wblock]

Does this mean I can get away with buying the cheaper WD Blue?

What is your data worth?

 

[Wipi]

What is your data worth?

Large photoshop files for printables (covers, pamphlets, flyers etc). Usually each file is around 1.5GB but sometimes a single file reaches up to 4GB. I changed my mind again and will go with 3x4TB WD RED RAIDZ1 so I may never need to buy more HDDs or expand the volume. I hope nothing goes wrong.

I will get Asus P10S-I which will fit nicely in a Fractal Design Node 304.