40
[Help] My first serious self hosted server (sh.itjust.works)
submitted 1 year ago by to c/selfhosted@lemmy.world
19 comments fedilink hide all child comments
 

Hello, I managed to get my hands on a second hand Proliant HPE server that I want to turn into a media server for myself (possibly family too).

I also have a bunch of drivers lying around in all different sizes. I want a good balance of security, backup and flexibility for the future. So hear my plan out:

  • Running Ubuntu server LTS on SAS 600gb disks (now it's in raid5 array with 3 identical disks but I probably want to change that and take out some disks from it for my data)
  • hardware raid 0 on the various single disks (with HPE smart array)
  • mergerfs and snapraid for a "raid" and backup (I read some information about it and I think for my use is the best option)
  • Headscale VPN (basically Foss tailscale implementation) for remote connection and mesh network
  • Docker with all apps

I'm no expert on servers or RAIDs or HPE. What do you think? I'm mostly worried about the hardware raid 0 + Snapraid, is it doable?

all 19 comments

sorted by: hot top controversial new old
[–] 35 points 1 year ago (6 children)

I used to have a boss that said that there are two types of IT admins, those that used hardware raid once and will never again, and those that haven't lost any data to it yet.

  • permalink
  • fedilink
  • source
  • hideshow 6 child comments
    •
  • [–] 13 points 1 year ago (3 children)

    Thankfully, we never lost data, but we have had the controller die a few times through the years on a few servers before. Thankfully we had great support and the part was same day delivered to us to replace and the 2nd time was a Dell server which we also get the part same day and a tech to come replace it for us.

    We avoided buying servers with this going forward. At this point, we really only have VM's and only a handful of physical servers anymore, so it's not even a concern.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] [S] 4 points 1 year ago (2 children)

    I read online that the controller failure is a thing and not that uncommon. I really don’t want to rely on HP hardware tbh.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 12 points 1 year ago (1 child)

    Parts fail all the time. The problem with hardware raid is you need a compatible controller or none of the data can be read even though it is still on the physical disks. Computer hardware is often only made for a few months before there is a new model and so you are risking that the manufacture really made the new model work with what you have. That is assuming the manufacture doesn't go out of business which could happen without warning. \

    Also, if hardware breaks that is often a good excuse to replace it - odds are better hardware is available for the same price and sometimes a lot less $ - with hardware raid you are stuck paying whatever price they charge.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] [S] 3 points 1 year ago (1 child)

    Well fuck me now I’m scared. So, for him the problem is hardware raid?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 6 points 1 year ago

    It all depends on the greed of the campaign. I worked in a campaign where it was considered normal to keep a degraded raid without repair. Of course, data loss is a normal story in such companies. The raid guarantees data security only when one disk is being pulled (except for some raids), so it also needs to be monitored and replaced. On the other hand, with proper operation, you probably won't lose any data.

    P.S. RAID0 - raid that can't be restored when degraded any disk in RAID. This is exactly worse choice for data save. STRIPE also writes blocks one at a time to the first disk and to the second, so that you would definitely lose exactly 50% of data blocks. Best choice raid10 for performance and raid5 if you need save money.

  • permalink
  • fedilink
  • source
  • parent
    • [–] 5 points 1 year ago (3 children)

    Just swap the raid controller for an HBA.

  • permalink
  • fedilink
  • source
  • hideshow 3 child comments
    •
  • [–] [S] 1 point 1 year ago (2 children)

    HBA is not real pass-through of drives, it creates a raid 0 for each one, striping them. If the controller fails, you can’t read anything.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
    • [–] [S] 2 points 1 year ago

    Thanks for the support, in the end I opened the server removed the raid smart array controller and connected the drives directly to main board. Checked within the BIOS and I see them perfectly.

    If you are looking for a HPE P440ar Smart array controller to buy hit me up :)

  • permalink
  • fedilink
  • source
    •
    [–] 2 points 1 year ago (5 children)

    When you say raid 0 on the data disks do you mean just having the disks present as single disks and not putting them into arrays? As seeing raid 0 and data storage makes me very nervous.

    And yeah I'd take a disk out of your boot array and then that into a raid 1 so you can use the extra for storage/ redundancy elsewhere.

  • permalink
  • fedilink
  • source
  • hideshow 5 child comments
    •
  • [–] 2 points 1 year ago (3 children)

    Ah gotcha. See if if your raid controller can be flashed/switched to IT mode(HP might call it something else) as then you won't have to deal with the raid controller's raid settings and doing anything weird. Then you can just rely on snapraid to manage the drives.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] [S] 1 point 1 year ago (2 children)

    Okay after a some search looks like I have 3 options:

    • stick to my original plan but consider multiple points of failure (controller and Raid 0 on single disks)
    • The “IT” mode for HPE is called HBA but as far as I read it just creates a bunch of RAID 0 single drives (lol). I also need to update bios firmware because I don't have this mode on my controller (huge pain in the ass)
    • Just switch from the Dynamic Smart Array RAID to legacy SATA AHCI and bypass hardware raid completely and switch off the controller. I need to reinstall ubuntu and headscale though :’(

    I think i will go with the last one.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 1 year ago (1 child)

    If you go down that route then look at zfs which will allow you some redundancy (like raid) raidz1 can tolerate 1 drive dying and radz2 can tolerate 2, but obviously it is not 100% though I have never had a drive failure corrupt a zfs pool in like 10 years, and replacing a drive is easy. This question has, I'm afraid, many answers. zfs on a 3 drive system would run nicely though, and zfs is supported by Ubuntu. But of course we are talking different ways to do things, and this is part of the fun of self hosting!

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
    • [–] 2 points 1 year ago

    I liked qdirstat

  • permalink
  • fedilink
  • source
    •