I am looking to begin self hosting, and has collected some of my older components together to see what I can do. I know it’s not much to start with, but I have 3 1TB drives that I plan/hope to use. My concern is RPM speeds, there is a 10k, 7.2, and a 5.4. Can I build 1 vdev out of these three?
I have been looking at various NAS software options, and trying to read and watch what I can about what is out there. At this point I’m between TrueNAS and Unraid. What I have gleaned is that Unraid mostly doesn’t care about what drive is added, just that the parity is the largest, which sound like my situation would would work. I have read some raid setups wont work well or at all with mixed speeds. The TrueNAS information I have gathered mentions the vdevs need similar drives. Is that in size only, or does RPM make a difference? Also, will different cache sizes, make a difference?
I would suggest going over the Hardware Guide section in the online TrueNAS docs. If you are using drives that old or small, I would be concerned for what else you are using.
The important part is making sure you use CMR type drives and not SMR. I don’t expect any of those drives to be SMR but it is worth a mention. You can also browse the Resources category for a lot of good info
SMR vs CMR ServeTheHome
TrueNAS Docs - WD Red SMR Drive Compatibility with ZFS
I second @SmallBarky’s suggestion of reading the Hardware Guide.
TrueNAS is very good, but, for using odd hardware or very new to NAS, it can be a higher learning curve than some others. This can bite people, (even badly), because they do something that seems okay, yet has risks they did not understand. Like using RAID-Z1 with lots of high capacity disks.
The SMR vs CMR issue I was partly aware of. I did a long SMART scan last night on the three drives, and googled their device model number vs smr or cmr, they all came back cmr. Though they are 10-15 years old, they came back as healthy, but I am still worried.
I looked through the hardware guide and don’t have any other concerns.
I also realize there is more to actual drive speeds than the spindles speeds, it is nice to know that TrueNAS should work for this mix. I will try downloading and installing this option, and “testing” before committing.
This is software raid so of course you can mix vdev speeds/latency.
Given non failing HDDs, the kernel and ZFS should manage everything without issues.
Just do it in mirrors regardless of new or very old hardware, to ensure at least one good copy.
(and then a 3rd local or remote HDD that gets updated with rsync or zfs replication daily etc).
For old HDDs, I torture test them, (full slow format under Windows) and perform transfer tests, short and long smart tests, before trusting or using them.
If you just want to find out what self hosting can do for you, i would try it with a virtual machine and virtual drives on proxmox.
Using 10-15 year old drives, even if they are working now doesnt mean they will be working in a year or two.
If you use the drives, I would use a 3 way mirror, and have a 100% up to date backup ready.
Also keep in mind that you need one SSD as a boot medium and at least one SSD to host the apps. Otherwise you would have to go into unsupported territory by partitioning the boot device or host the apps on HDDs- which will work but be very slow for random access small files like a lot of apps create.
One odd thing with generic HDDs. Meaning HDDs that are not normally used for server or NAS functions.
This odd thing is TLER, Time Limited Error Recovery. Seagate calls it something else, but the concept is the same. Basically during sector errors, desktop and laptop drives will try extensive attempts to recover the sector. Multiple reads, applying the error recovery code, etc…
Sometimes TLER is set to over 1 minute for desktop and laptop drives. But, NAS and server drives tend to be 7 seconds or less. This is because redundant data is available via various RAID schemes, (well, except RAID-0). In fact, some RAID schemes, both hardware and software, can time out if the drive takes too long to respond for a read request. This can toss a drive out of the RAID group, and list it as failed.
For example, a co-worker had a Linux software MD-RAID set on a home NAS server which occasionally tossed out a drive. And not always the same drive. They always re-synced properly, indicating they were still usable. I suggested NAS specific drives, like WD Red, (this was before WD polluted the Red with SMR drives…). He did not have the problem again.
Today, I would have suggested checking to see if the TLER could be changed. And using SMART for verification that the drive is good, (just loosing a sector).
To be clear, occasionally a server class HDD has TLER set high. Not sure why, but I think someone here in the forums has seen it in the last few years.
