Hacker News Reader

Comment by ink_13 | Hacker News Reader

I can see the attraction for enterprise-level storage, where a single chassis might contain anywhere from 20-90 drives, but my thought on seeing news like this is that surely we're getting to the point where resilvering times will make small arrays infeasible from a reliability standpoint.

My home NAS only has four drives in it: if one 20TB drive dies, how long will I have to run on three, even if I have a spare onhand? Will they all survive long enough to rebuild? Or am I going to have to run some RAID-6/RAID-Z2-like array just to ensure durability?

View on HN

Standard reminder to home / small biz RAID users: it is an availability solution, not a replacement for backups. It lets you have fewer service outages connected to disk failures. There are many more ways to lose your data than physical media failures.

100%. I much prefer snapshots. All of my data loss situations have been in accidental deletion / formatting situations which raid does nothing for.

I can’t see any use for raid as a consumer other than raid 0 for speed.

Snapshots also aren't a replacement for backups, as they reside on the same media/device.

They can be if the snapshots are copied to another device though

This is for enterprise level storage. As you've said the resilvering time is too long unless you're using something like object storage with a whole cabinet to eat the reduced EC shard availability. But the good news is that you have access to enterprise level storage for your data that you want to make sure is durable. Use cloud storage.

If you have a spare onhand and a small array you should just put it into the array and make things massively more reliable.

If you have to go get a new drive, then reliability won't change a huge amount based on rebuild time.

it's still true that once you lose a drive you want to pop in a new one and resilver, it just reduces the risk to have 2 spares/parity drives instead of 1.

my array is relatively small and it still takes like 24 hours to resilver one drive. pretty brutal.

my current risk mitigation strategy is to have one spare in the array, and then have one 'warm spare' in the nas that is not part of the array and is instead a target of scheduled backups of the whole array every week. that way if i somehow lose another drive in the array while resilvering is in process i'm not totally screwed. (if my NAS could hold more drives I could do both, but more bays is way more price...)

naturally I also backup to the cloud. but you gotta have a local solution too.

Your usage is different than mine: I have a 18TB disk in my desktop tower so all my data is always available and have a second tower with 10x2TB disks in raid 6 for backups (all my old disks).

I switch on the second tower once a week and rsync everything from my desktop.

I don't have a solution if both fail at the same time but the probability of it happening is low enough that I accept the risk.

Ideally I would need a third backup rig somewhere outside my house, but I can't afford this level of safety.

Why not use erasure coding like in Minio instead? Software more reliable long term than hardware anyway

These drives would be particularly bad at resilvering as they are cmr drives.

Been there done that: never again.

>These drives would be particularly bad at resilvering as they are cmr drives.

Are you mixing up CMR and SMR drives?

Reply on news.ycombinator.com