I saw these very cheap disks on ebay, $37 for 1.2Tb ea, I need help figuring out the best way to use them, or if I should just throw them away. I just want to host some websites and experiment with kubernetes/docker/have some fun doing self-hosting projects.
Storage: 10x 2016-2017 Dell 1.2TB 10K SAS 2.5" 12G SAS HARD DRIVE. ($0.0316/Gb).
Available: 12 more SAS 2.5” bays in front, 2 more in back, plus 6 PCI gen 3 x16 slots.
ZFS seems like the safer option, but I’m worried will it be too slow? The hardware raid seems tempting but I’m not sure if it’s a genie in a chip.
Is ZFS or hardware raid the correct move here? Raid 10?
P.S. Apologies if this is the wrong category for this, I am new here.
my preference would be ZFS because of zvols and all that - but you need to mention how much RAM you have in order to get a feel of whether or not it’ll become a bottleneck. but as CERN has persuaded me if you have an HBA then you should prefer that for peak performance…(and maybe use ZFS on top of the single “disk” if you really want).
ZFS doesn’t really like being run on HW-raid controllers. Some controllers use a abstraction layer and don’t allow direct disk access (which ZFS employs to do its scrubbing and health checks), secondly you ignore the biggest advantage of ZFS: portability of the whole pool(s) ontop of adding another point of failure.
For slowness or RAM usage - 12TB of total storage space should be a non-issue. 8GB - 16GB RAM should be enough if you don’t run deduplication or sth similar or run other demanding services (if you are low on RAM). I run a ZFS volume of 4TB on a Skylake i5 6500 (Lenovo Tiny).
There is nothing wrong repurposing old, working drives. On my server with 25 bays, I simply threw 5x 4TB Desktop SATA HDDs in a ZFS1 pool to make them do sth usefull besides collecting dust.
ZFS all the way. That H730 with the battery is fine for plain RAID, but once you go hardware RAID, ZFS can’t see the individual drives – no scrubs, no self-healing, no snapshots, no send/recv. For self-hosting and Kubernetes, ZFS snapshots alone are worth it. Those 10K SAS drives will be plenty fast for hosting websites and containers – just add an SSD mirror later if you need more speed. With 10 drives, go RAIDZ2 for the two-drive redundancy margin, especially since those drives are from 2016-2017.
One important thing: flash the H730 to IT mode first, or at least make sure each drive is passed through individually. If it’s stuck in RAID mode, ZFS won’t work properly. And check the SMART data before you trust them with anything – power-on hours and reallocated sectors will tell you if they’re near end of life. At $37 for 1.2TB it’s a great deal, but expect at least one to fail in the next year or two. RAIDZ2 gives you room to replace them one at a time.
As @Dratatoo (… and @linkdford ) said, just make sure your storage controller can give you direct access to the disks without any caching or other shenanigans in the way that would make ZFS cranky. Usually that’s called IT mode, but I’m not as familiar with the old Dell controllers.
Thanks for the replies! ZFS sounds like the winner. While I’m not excited about the potential performance hit, the others pros seem worth it.
@Dratatoo I have 64GB of DDR4, upgradable to 384GB. When you say 8-16GB should be enough, do you mean for the whole system, or should I be expecting ZFS to consume about that much?
@linkdford thanks for calling out the SMART data check. You are right when H730 is in RAID mode, my 10 drives appear as one virtual disk. When I set it to HBA mode I can see each drive.
After posting this last night, I found this Level1Techs video from about 10 months ago, “Talkin’ ZFS With Allan Jude!” that had a number of insights.
Really slow IO to VMs can be caused by using wrong sector/record size
Database record size should match the page size of the DB (eg PSQL 8kb, MySQL 16kb)
VMs - Allan recommends 32k or maybe 16k.
So great, ZFS looks like it can be tweaked to get better performance, but I also open the door to more readily shoot myself in the foot by misconfiguring it. Hooray!
For this kind of home lab, I would look at what you want the storage to do before choosing between the H730 and ZFS. Since you are mainly hosting websites and experimenting with Kubernetes, Docker, and self-hosting, you do not necessarily need to get the maximum performance from those 10 drives.
The 1.2 TB 10K SAS drives are still useful for a lab. I would not throw them away simply because they are older.
If you want to use ZFS, the main consideration is that the H730 should not sit between ZFS and the drives as a normal hardware RAID controller. ZFS works best when it can directly manage the disks. You would normally want the controller in a suitable HBA/pass-through mode, if supported by your exact setup.
For 10 drives, a few layouts are possible:
• RAIDZ2: Good balance of usable capacity and protection against drive failures.
• RAIDZ3: More protection, but less usable capacity.
• Two RAIDZ vdevs: Can provide better I/O characteristics, but uses more disks for redundancy.
• Mirrors / RAID10-style layout: Good IOPS and easier rebuilds, but only about half the raw capacity is usable.
For your workload, I would lean toward ZFS with redundancy rather than hardware RAID, particularly if you want to learn ZFS and use features such as snapshots, checksums, scrubbing, and easier storage management.
Performance-wise, 10K SAS drives will obviously not feel like modern NVMe storage. However, for websites, Docker containers, Kubernetes experiments, backups, and general self-hosting, they can still be perfectly usable. Your biggest limitation will generally be random I/O rather than CPU power.
One thing I would definitely do before trusting the drives with anything important is check their SMART/health information, power-on hours, error history, and bad sectors. Drives from 2016–2017 have already had a lot of operating time, so their condition matters more than the low purchase price.
For a lab, I would also avoid putting irreplaceable data on the array just because it has redundancy. RAID/ZFS redundancy is not a backup. Keep important data somewhere separate.
So, for your particular R730XD, I’d treat these disks as a great homelab storage pool, test their health first, and use ZFS if your goal includes learning and experimenting with modern storage management. If you primarily want simple, predictable block storage and already know the H730 well, hardware RAID with RAID10 is also a reasonable approach.
IMHO, these drives are not worth the power to operate them 24/7. You should be clear what you want to use these drives for.
You’d need at least one drive of redundancy, that would leave ~10TB usable storage at best.
You can get a single (used/refurbished) 10TB HDD for less than $370 you paid for these SAS drives. It probably is as fast as the whole 10 drive setup combined consuming 5W instead of 60+W. Less hassle to setup and operate.
OTOH, if you don’t have a lot of experience in building and using (Z)RAID configurations these drives offer a great learning experience.
Turn 'em on a couple of hours after work, shut them down when not in use.
There is lots of fun to have with this setup to play and learn (well, homelabbing).
If you’re looking for more than part-time use, I’d go with the third option (“throw in garbage”). These drives have been obsolete technically for decades.