35000 power-on hours. SMART still reports it as OK.
Time to figure out how to rebuild a RAID 5 array. The other two drives are probably nearly cooked too, but I have plenty of spares that I got for free.
Perfect occasion for one of the remaining drives to die during resilvering. I hope your backups are up to date?
No backups. For important documents and photos, that was the backup and most of them have copies on my PC. The rest is easily replaced. I knew what I was getting into, and that the free, decommissioned hard drives with 20-30 thousand hours on them were a lit fuse.
Looking into making my own NAS, where’d you get these?
Benefit of my job: I get access to the scrap pile. I don’t know any reputable used/refurbished sellers.
Maybe look into the 3-2-1 backup scheme while you’re at it.
I know what it is, and I ensure compliance at work (I’m a sysadmin). At home, it’s less about best practices and more about what hardware I can afford. Manufacturers tend not to offer regional discounts. A 2-2-0 scheme is better than nothing at all.
“Yeah, I just saved them to the NAS earlier today before it started making noise”
- User about to have a bad day, possibly me in the future…
My predecessor at work had a “backup scheme” where each week a full copy of important VMs’ virtual disks would be pulled by a backup VM. Two issues with that. One, the VMs were not powered off and nothing ensured that the disks were synced. Two, the backups were made onto the same physical host with no replication or high availability beyond RAID 1.
Yeah, at the moment I am building a home server running TrueNAS, it is the only backup I will have, but it is far better than having everything on a single HDD in my computer…
I need the final two HDDs to set up my raid and start using the server.
I have a question about my raid though.
Is it better to run it with four data drives, one parity, and one hot spare, or should I run it with four data drives and two parity drives, and no hot spare?
I am running 8TB Seagate Iron Wolf Pro drives, from (hopefully) separate batches.
The CPU is an AMD Ryzen 4600G and I have 32GB ram in it.
I want to add another SATA controller and add two SSDs, one for L2 cache, and one for applications and VMs, as well as add a low profile GPU for transcoding and later a 10G NIC
This is something the people of [email protected] are better suited to answer.
In my personal opinion, for most home servers, the double redundancy of RAID 6 is more valuable than having a fully rebuilt array as soon as possible. If one member of a RAID 6 array fails, you’re still at an effective redundancy of a RAID 5. If one member of a RAID 5 array fails, you have zero redundancy until the hot spare is rebuilt.
If energy consumption is also a factor, it’s worth keeping in mind that a hot spare can be powered down by the controller until it is needed.
I’d personally go with RAID 6: 4 data + 2 distributed parity. That was the plan for my server too, but the motherboard only has four SATA ports and one had to be dedicated to the OS SSD.
Thanks, that makes sense, I picked the server motherboard specifically because it had six sata ports, they all can be used even with the first M.2 slot used.
I still want to add an L2 cache and an SSD storage pool for VMs and Apps. So I’ll need another controller card.
One annoying thing is that the server constantly waits at the boot screen untill all drives have spun up despite not having anything to boot from on them…
Also check temperature, if the noisy boy is also hotter then it’s neighbors… That’s a strong indicator
You’re probably already working on replacing it, but SMART returning “okay” doesn’t rule out a failing disk. What you want to look for on SMART tests are the RAW value of reallocated sectors. Anything over 0 is a bad sign.