Collateral risks of running unreliable drives?

[Dice]

Hey!

When I got interested in FreeNAS, one of the key selling points was the idea that drives not reliable enough for NTFS data keeping could be put in a separate 'trash pool' to extract the last bits of life out of them.
I still have approx 5-8 drives that would be suitable for such a pool. Yet I've not come around to get it going. Partly due to lack of space in current case.
I've not given up on the idea. I'm considering using it for either a landing zone type of storage area, or live backup of the more important stuff.

I wonder if there is any sort of overlapping or collateral risk of running a pool of unreliable, smart-error-piling drives along side a pool of fresh drives?
That is, if a catastrophic failure of the unreliable drives would in <ANY> event (what sort?) result in damage or putting the fresh pool at risk?

 

[wblock]

I'm curious about what kind of data would be stored on such a pool. ZFS counts on at least some reliability in the pool. If the drives are already so near failure that they can't be used alone, the odds of losing data due to the increased wear of resilvers goes up: one drive fails, other drives fail due to increased activity of resilver onto replacement drive, additional drives fail before resilver completes.

Using this as some kind of experiment I could see. I would not put this sort of a bound-to-fail pool in an existing machine due to the added power draw, heat, and possibility of some kind of hardware failure shutting down the rest of the machine.

Where was this selling point brought up?

 

[Stux]

I use pools like that as backups.

Only risk I can think of (as in cross contamination) is if swap is in use.

 

[Ericloewe]

I mean, RAIDZ3 can do miracles, but at some point you have to ask what you stand to gain from said miracles.

 

[Johnnie Black]

My backup FreeNAS server works as a third backup, so I use it with drives that have seen better days, most have just some slow sectors but there's a couple with a few pending sectors, still going strong and already replaced some disks and no issues during re-silvering.

 

[rs225]

All drives are unreliable drives. Since you know ZFS, the concerns would be the swap and hangs/timeouts for reads/writes to problem drives. I think only the swap would be a risk to the other pool.

 

[wblock]

All drives are unreliable drives.

Well, yes, although it varies in how unreliable they are. At some point, redundancy and resilvering will not be able to keep up with the bathtub curve failure rate.

 

[Dice]

Where was this selling point brought up?

Nowhere. Rather as one of my own conclusions to the appeal of using a self-healing file system, with software raid.
As opposed to other solutions I've been fiddling around with pre-FreeNAS.

Using this as some kind of experiment I could see.

That's where Im at right now - but in contemplation regarding the potential SWAP problem.
That threw me off a little I must admit.
Glad I posted.

As for now, I recon it would make the most sense to put the sketchy drives in their own machine that would be used for backup replication.