backups > raid
do what you gotta do
hardware raid is fine and has the benefit of being relatively simple/cheap to implement. You can get an old dell perc card for cheap and they will work great without the added overhead of something like ZFS. Obviously you lose out on some of the featureset. Mirrors are good enough for the average home gamer.
reee, I don’t game, I run loonix!
get 2 externals in case the cat fucks one up
Inefficient backups are better than none.
How much data do you currently have, and how much data are you planning to store (and what kind)?
atm ~1.4 TB. Video, Pictures, Music, Books, code.
But I also wanna store some of the family stuff, like their photos and what not.
I would say a 4tb with a 4tb for backup is fine. When you outgrow your 4tb storage (since you have a backup) you can easily build a raid5 or whatever really at a later time (personally run a mdadm raid6 too).
So once you need more storage you could buy a 8tb backup drive and another 4tb one copy the backup to the 8tb one and build a raid5 from the 4tb drives.
One thing you can do in the meantime is encrypt/compress your most important things (family photos, financial records, illegal memes) and upload it to some of the free cloud services available. That’ll give you some good extra redundancy.
i would say yes. Built a NAS recently for about $400 all in. 3x 2TB HDDs in RAID-Z(5) for a total of 3.5TB capacity. Not a huge capacity, but I wanted reliability and redundancy in case one drive fails. (Also, I don’t need crazy high capacity).
It depends on your use case and personal preference.
This:
I’d be going with a minimum of 2-3 appropriate size SATA drives to mirror.
If you find performance is inadequate (i don’t think you will, as per above - you will be NIC limited), step up to 4 drives for a striped mirror set up (2x mirror vdevs in your pool). ZFS will serialise random writes and cache reads, so again, performance for single/small number of users should be plenty.
Sizing is a case of either needing more spindles for speed (you won’t) or concerns regarding rebuild time. Bigger drives take longer to rebuild (ZFS will only rebuild what is consumed, but presumably if you buy a big drive you will eventually go close to filling it).
With 8 TB drives, i’d be tempted to get 3 of them (3 way mirror) in case of failure during rebuild, but YMMV. 2 will likely be just fine. Probably better to back up your super critical shit off-site anyway, if its more a case of losing the non-critical bulk of the data being inconvenient as opposed to life changing, i’d consider a 2 way mirror to be good enough.
If a 2 way mirror isn’t enough then off-site backup is probably essential anyway. RAID/ZFS isn’t a backup of course, but it will maintain service in the event of a failure.
Not a fan of cloud, but yeah encrypted tarball could be a thing.
If you run something like freenas i’d suggest setting up owncloud/nextcloud and working out of your synced folder for really important stuff.
If your NAS breaks, your important stuff will be synched to your workstation PC, and vice versa.
Combine that with ZFS snapshots on the owncloud/nextcloud dataset and you’re safe from synchronisation fuck up, if that was to happen, also.
Also, just on backup.
If you were to go for say, 3x hot-plug drives in a 3 way mirror, you could perhaps remove one periodically to use as an off-site backup. Plug it back in , let it re-silver and then remove either that one or a different one next week/month to keep off-site.
You’ll still have hardware redundancy with the 2 drive mirror that remains, and if you have a failure, just pop the backup drive in for a fast rebuild as it will already be mostly synchronised from before you popped it out. ZFS only rebuilds what is missing.
If you have total system failure (house burns down, etc.), you’ve got your 1 off-site drive with (most of) your data on it that you can import into anything that can run ZFS to get your stuff back.
No need to fuck about waiting for backups to happen, just pop a drive out. Mirrors are a wonderful thing 
. Sometimes, the best/most convenient backups are the ones that require no time or effort to do…
Alternatively, zfs send/encrypt your stuff to a cloud provider.
that sounds like very good idea, maybe one day when I am not poorfag but wealthchad I’ll do it 
There’s some refurb hgst Enterprise drives out there with 3 year warranties.
4tb for ~$70
from experience seagate are garbage, they work fine … till they don’t most of the time there no warning, SMART was displaying the drive as healthy yet they died at least WD you usually get a heads up before failure
when you buy drives for nas just remember the more drives you have together the more they affect each other
1 to 4 drives wd blue are fine
id start using red at 4 but there not a must till 8 + to be honest
20+ use red pro or enterprise
as a permanent solution external drives are too vulnerable, and usb is wayyy too flaky for reliable long term usage, those are best for offline backup not running storage
Ruffalo the price is not the same everywhere for example
WD My Book 8TB USB 3.0 External Desktop Hard Drive WDBBGB0080HBK is $320 in Australia you shouldn’t assume the op is in the us
anyway i’m gonna go in the opposite direction from everyone else about drive sizes when everyone say buy big and ruffalo says one should never buy bellow 6tb in 2019, i completely disagree
i have probably 120hdd running here and none is above 4tb the reason is sata drive throughput isn’t fast enough
in raid drives always rebuild slowly… especially if the system is in use it’s not rare for a 2 tb to take 24 hours to rebuild while this is the main reason why bigger is not always better it’s not the only one
with big drives you end up with a small number of drives in your raid if you only have 1 parity you are more then likely to lose everything as a second drive fail during the massive amount of time a 8 tb will take to rebuild or if you don’t change the drive right away and all your drives have the same age it might even fail before rebuilding even start
next lets imagine you have 2 8tb drive in raid 1 the drive read and write speed will just abysmal, if your main system has a ssd try doing to a hdd only system for a few days, you might be able to have 1 maybe 2 user at most or you will have sporadic performance
raid 5 and raid-z1 have the same single parity issue despite needing 3 drive minimum but slightly better throughput… sometimes, overhead is troublesome at times
if you look at raid 6 or z2, it require at least 4 drives (6 or 10 is better)
the benefit is simple better read and write speed, 2 parity helps a lot when rebuilding
then you have what i use, raid-z3
my chassis are: 24 drive bays with 2 pools of 11 live drives and 1 hot spare
the main pools are all made of 2tb drives and the pools where they duplicate are 4tb drives
zfs memory usage isn’t bad really as long you don’t use de-duplication
running 24 x 4tb system (11+1 x2) on 8gig of ecc ram works well
meanwhile with dedupe well that’s a mix bag of… the 1gig per tb is just a guideline some time it’s not enough from experience z3 with 11x2tb+1hot spare x2 give about 31tb of usable space but the system would never work properly with 32g of ram it just wouldn’t
You don’t use them over USB, you crack them open, pull out the WD blues inside, and use them in your NAS.
backblaze used to do that and they found that the process actually make the drives die a lot more it’s actually fully documented and not recommended
you shouldnt assume we dont know where gek lives
For the record the easystores are the ones you wanna go for if you’re trying to shuck on the cheap.
good come back but it’s just better to generalize rather then localize based on assumption that’s all i meant plus other might stumble on the conversation for there own needs too the wider scope is a good thing over all
Source link, please. Shucking is extremely common and I’ve done it for years myself.