@zirias

@parvXtl

One of the ways I help avoid failures with raidz1 is by replacing HDD when they get to around 5 years old, or they start showing errors. Also, monthly scrubs helps ensure that the drives are getting fully utilized regularly which helps keep things healthy/accelerate detection of failures.

#resilver #FreeBSD #RAID #ZFS #raidz1 #backup

@david_chisnall My pool isn't for a NAS but basically for "everything" I need, which includes serving files, but lots of other things (routing, firewall, internal MTA with mail storage, building base and ports, a windows server 2020 for work, and so on ...). Of course I avoid IO-heavy stuff during resilver, but it's still a busy machine, so I guess a "balanced" default tuning is actually what I want 😉

Thanks for explaining how ETA of #ZFS #resilver ends up so "random" 👍

For whatever reason, #resilver on my #FreeBSD server slowed down continuously, an initial ETA showed around 4 hours, by the time I left my desktop, it already showed around 10 hours. But hey, it finished without errors, so, all fine 😅

BTW, I really don't get why recently, you read a lot of stuff about how #RAID-5 (or the #ZFS equivalent #raidz1) was super dangerous and you should never use it 🫨

What's certainly true is: With larger pools and larger individual disks, the risk of a second disk failure during resilver significantly increases. But then, there's no "risk-free" storage, so a #backup is always a *must*.

What's also true is, raid-5/raidz1 is still the rendundancy scheme with the least storage overhead for most scenarios (3 and more disks). And of course it still reduces the risk of a failed pool. This pool here has only one of its original disks left, I didn't need my backup so far. 🤷

So please move the discussion of RAID back to a sensible base. The scheme/level you choose is always a trade-off between the cost (overhead) and the amount of risk reduction, and that's pretty much all there is....