So I’ve been running a home media server with the arr suite on it, but it only has about 2 tb storage. The storage has been full for a couple months now which has rendered my server useless. Now I’ve bought a used nas with 32 tb on 7 disks. I’ve been advised to set it up in RAID. I’ve learned how to do that in OMV and it’s setting that up now.

It takes a while though, so that leaves me some room to think. Why do we do this? I understand that it gives me more unified storage, but I also want to run other things on this machine. Like k8s to deploy my own applications and play around with devops stuff. Or even just regular file storage and photobackup next to the films and series. I thonk I would like to keep those things seperate though, is that still possible with RAID? Or is it better to put 5 disks in RAID and keep the other two for other stuff? I think my question is, how do you use RAID “practically”? How do you store different things on there? Do you compartmentalize the disks again, or do you just make different folders?

  • MuttMutt@lemmy.worldEnglish
    4·
    3 hours ago

    Software RAID, yes. Hardware RAID, NOOOOOOOO!!!

    Without knowing what the hardware is in the system I need to stress that before getting started.

    Software RAID is controller independent, if the MoBo dies tonight you can get a replacement of whatever that is cheap or even grab a HBA (Host Bus Adapter) and hook the drives up. Hardware RAID requires an identical controller with the same version of firmware in most cases. In a home environment hardware RAID is a major pain, most corporations have spare hardware on hand if there is a failure so it’s not as much of an issue. Also JBOD (Just a Big Old Drive) is not the same as a HBA so using a hardware RAID controller with that turned on for a software RAID is asking for issues. For software RAID in OMV you will be looking at something like BTRFS https://en.wikipedia.org/wiki/Btrfs or using a plugin to add ZFS https://en.wikipedia.org/wiki/ZFS

    I come from a FreeNAS/TrueNAS background using ZFS and SAS2008/SAS3008 HBA’s.

    There are different version of RAID array’s, you need to look them up and decide what works best for your data. What works for me or anyone else is irrelevant. No RAID is not a substitute for good backup policies, but it saves a potential TON of work doing a restore and the headaches associated with it. https://en.wikipedia.org/wiki/Standard_RAID_levels

    RAID 0 or Mirrors is great for data integrity. You could put all 7 disks in a RAID 0 or mirror configuration and it would take 7 disk failures to lose your data.

    RAID 1 or Stripes is great for speed and storage capacity, the data is split into chunks and with all 7 disks you would have maximum capacity of the total sum of all the disks, one disk fails and all your data is gone like a fart in the wind.

    RAID 0+1 or Striped Mirrors combines the two above concepts. In some ways you get two disk failure and in others you don’t. Take 4 decks of cards, set two side by side, those are your mirrors. Take the other two and put them above the other pair, those are the stripes. Lose two mirrors and the data is gone, lose a single mirror in each stripe and the data is safe. You can mirror and stripe to your hearts content, want 3 mirrors with two stripes it takes 6 drives.

    RAID 5 or RaidZ 1 gives you one disk failure redundancy, how it functions depends on the raid controller or software utilized but basically data is split in to chunks and you have the ability to replace a single disk and fully recover. If a second disk fails your data is gone again. Some people consider this dead as the odds of a large drive having an issue during a rebuild is higher. https://www.servnetuk.com/insights/ure-is-raid-5-dead

    RAID 6 or RaidZ 2 adds a second disk failure redundancy. If you want maximum space this is what you should probably use with failure redundancy. It’s not super fast but unless you need the speed for anything in particular it’s a good trade off. With 7 drives this is a good option.

    RaidZ 3 Adds a third disk failure redundancy. This is what I use for most of my pools. When rebuilding you are stressing the disks to the maximum and that is when you will find a second issue pop up. In a RaidZ1 your data is gone, in a RaidZ2 you lost your last spare. In a RaidZ3 you still have one failure that will save your bacon.

    I have multiple pools. I have the majority of my regular data on a RaidZ3. I have a couple striped mirrors for things like VM’s, Docker containers, and things I want a little more speed on.

    You can also do things like RAID 0+5 or 1+5. In ZFS this would be called RaidZ1 with two vDev’s.

    I use TrueNAS because of ZFS, BTRFS is still a work in progress in comparison.

    I imagine about now you just about had your head explode, it’s tough to make this decision because it puts you down a path. I started when BTRFS was an infant, I stick with ZFS/TrueNAS because I know how things work and it suits my needs. You need to decide what will work for you and you may find that OMV isn’t the best path going forward. Take the time to really make a decision about what your needs are today and what they will be in the future 5 to 10 years plus down the road. Don’t just say, well (random person on the internet) says this is best so it’s what I will do. Every decision is about tradeoffs and compromises.

    Once you make a decision one of the best things you can do is do a burn in on the drives. I do it with BadBlocks and run smart tests. I just got done with a batch of 8TB SAS drives and found 4 with issues that were replaced before I even put them in to production. If a drive survives a torture test early on it’s less likely to have issues later on. I upgraded from a pool of 4TB drives and out of 11 drives I had 1 failure. When you get to the point where you are thinking about expanding your pool with new drives you will want to buy from a couple different sellers, at different times a week or two apart, or different manufacturer’s.

    One other thing I will add… Toshiba is verboten in my world. They don’t replace, they give you a check and you have to buy at current prices and it’s more of a PITA than I want to deal with. BackBlaze is also a great place for information about drive reliability. https://www.backblaze.com/cloud-storage/resources/hard-drive-test-data