Need help setting up HP Microserver N40L

• 

  _Bruce_ on 17/11/2012 - 08:36

  I think RAID 5 usually the best solution. Consider making it with only the 3TB drives and expanding with more as it will give you better capacity. Although it seems silly you can make RAID 5 using only 2 drives, which means you can throw more in later (you can even make it with one).

  I generally find Ubuntu one of the easiest for non-unix people to pick up, so unless you know someone who is familiar with something else, probably best to start with that.

  As for limiting access, this is about users rather than computers (set up a user/password needed to access file sharing).

  • 

    timmy_tam on 17/11/2012 - 09:34

    thanks very much. i'll give it a go. i take it you install ubuntu onto one of the drives and raid 5 the remaining?

    • 

      _Bruce_ on 19/11/2012 - 07:32

      You could do that, but I wouldn't. You can install onto RAID, but it is a bit more complicated. Alternatively, partition off a few gig on each drive, and install to one of those, and use the rest of the drive in RAID.

• 

  DARK KHAOS on 17/11/2012 - 09:59

  I installed freenas on one of those office works usb sticks.
  the microserver has an internal usb port on the motherboard specifically with this in mind.

• 

  tonyjzx on 17/11/2012 - 11:19

  ubuntu is way overkill

  freenas or nas4free would be better

  • 

    _Bruce_ on 19/11/2012 - 07:44

    Fair point. I've never used these so didn't want to recommend them.

• 

  inm8 on 17/11/2012 - 13:57

  +2

  This post on OCAU is well worth a read: http://forums.overclockers.com.au/showthread.php?t=833690

  If you want a web GUI FreeNAS is the way to go. If you'd like to your hand at the command line (so much simpler than any GUI could be once you know your way around, IMO), then get a Solaris distribution like Openindiana and follow one of the many guides on the internet for setting up a ZFS fileserver with SMB enabled.

  As far as your disk configuration goes, don't be tempted to use Raid-5, Raid-Z, or for that matter, anything other than mirroring, for the slight increase in capacity. Remember storage is cheap and is only going to get cheaper. For most of us nowadays, at least some of our data is important. My suggestions for your disk configuration, assuming you go the ZFS filesystem route (FreeNAS or my Solaris suggestion):

  Use three drives to begin with:

  • Two drives in a ZFS mirror
  • One drive as a hot spare

  This has a number of benefits over Raid-Z. By the way, Raid-Z is the ZFS implementation of Raid-5:

  • Two drives need to fail for you to lose your data.
  • Your read speeds are faster because you can read from three drives at once.
  • Storage capacity expands organically over time: Lets say you start off with three 1TB HDDs. Your server will have just under 1TB of usable storage. When the first drive fails, you replace that 1TB drive with a newer, faster and cheaper 2TB drive. The mirror is rebuilt on the new drive and you've still got 1TB of available storage. When the other two drives have eventually failed and are replaced with 2TB units (or larger), the storage pool (your servers available storage) will automatically grow to 2TB. Given that storage gets cheaper over time and your also likely to collect more data over time, being able to easily grow your pool incrementally is invaluable.
  • If you're really strapped for cash then you can do this with two drives to begin with and have the same size storage pool, with reduced redundancy (1 drive failure versus 2) and reduced read speeds (still faster than Raid-Z/Raid-5 in most usage cases).

  If you need any help just ask.

  • 

    _Bruce_ on 19/11/2012 - 07:43

    Sorry, don't agree about RAID 5 Vs 1.

    Two drives need to fail for you to lose your data.

    If you want this use RAID 6. Better way to do it anyway, as this rule will scale more consistently (2 failures is pretty much the definition of RAID 6).

    Your read speeds are faster because you can read from three drives at once.

    Same with RAID 5/6. Actually, these are faster when you end up with more drives in the same array.

    Storage capacity expands organically over time: Lets say you start off with three 1TB HDDs. Your server will have just under 1TB of usable storage. When the first drive fails, you replace that 1TB drive with a newer, faster and cheaper 2TB drive. The mirror is rebuilt on the new drive and you've still got 1TB of available storage. When the other two drives have eventually failed and are replaced with 2TB units (or larger), the storage pool (your servers available storage) will automatically grow to 2TB.

    So I have to wait for all three drives to fail before I get any more space? And even then I only get as much as the first replacement drive? And there is no way to just add more space by adding another drive?

    Given that storage gets cheaper over time and your also likely to collect more data over time, being able to easily grow your pool incrementally is invaluable.

    Hasn't changed much for years now. And probably won't again until SSDs overtake magnetic (which to be fair isn't too far off). As for easily growing RAID5/6 is so far ahead here there isn't really an argument.

    If you're really strapped for cash then you can do this with two drives to begin with and have the same size storage pool, with reduced redundancy (1 drive failure versus 2) and reduced read speeds (still faster than Raid-Z/Raid-5 in most usage cases).

    … or do the exact same thing in RAID5. And this gives you all the above advantages when you expand. Infact, if you have a 2 drive RAID5 array, the performance will be exactly the same, only when you have at least 3 drives does it need to calculate the XOR and slow down write speed (but not read).

    • 

      inm8 on 19/11/2012 - 09:32

      I think it's a good thing there are so many ways that we can choose to approach the same problem. Some reading material which helped me come to my decision:

      RAID-GREED and why mirroring is still the best: http://constantin.glez.de/blog/2010/01/home-server-raid-gree…

      3-ware technical brief on Raid-6: http://www.3ware.com/products/pdf/RAID_6_techbrief_112906.pd…

      Why RAID-6 stops working in 2019: http://www.zdnet.com/blog/storage/why-raid-6-stops-working-i…

      Raid-6, do you really want it: http://www.techrepublic.com/blog/datacenter/raid-6-do-you-re…

      • 

        _Bruce_ on 19/11/2012 - 09:42

        Some of the technical details in the first article are not quite correct. The others are not really talking about home use.

        If you are planning a server for a data center and you know exactly what it will do and when it will be decommissioned, sure there may be a better option.

        For flexible home use I really cannot recommend anything other than 5/6.

        If you think you need better reliability than 6, a home server is NOT the right option anyway.

      • 

        inm8 on 19/11/2012 - 09:48

        Better reliability with a simpler, easier to understand form of redundancy (ie: mirroring) makes all the sense in the world for home use if you ask me. Each to their own though.

      • 

        tonyjzx on 20/11/2012 - 11:39

        its sorta a waste of time going into these details for the typical N40 user

        problem is if you understand all that you dont need to ask for help

        you'd be better off with raid1 or something simple the avg. home user can understand

      • 

        inm8 on 20/11/2012 - 15:04

        Agreed. Raid-1 is a form of mirroring.