Help me to rebuild my NAS

    This site uses cookies. By continuing to browse this site, you are agreeing to our Cookie Policy.

    • Help me to rebuild my NAS

      Hi everyone,
      I built my first NAS 2 years ago and after those two years and the help of many of you I think I'm ready to rethink my NAS from the scartch.
      Right know I have:
      -a 60GB ext4 SSD for OMV
      -1 RAID1 of 4TB as EXT4
      - 1 ZFS Mirror of 6TB

      Right now I still don't have 4TB of data so I'm still in time to rebuild everything from the beginning.
      What I need is a NAS that will work flowlessy without particul problem or manteinance. I will use it mostly for backup my photo (I also have an external HDD for them) and as a Media Server with Plex.

      This is my idea:
      - SSD: give OMV a 20GB partition, create a 20GB partition for settings of application. I will install them using docker and a 12GB swap-partition so that I can hibernate my NAS;
      - Instead of RAID1 and ZFS Mirror I was thinking about a sync job that will sync from 4TB1 to 4TB2 and from 6TB1 to 6TB2 . I think that this way it should be easyer to manage my hdd and in case I delete something for error I will still be able to recover it.
      - Every partition will be BTRFS since I read on Github that OMV5 maybe will drop the support to ext4;

      This will be the final configuration:
      - SSD: 20GB BTRFS for OMV
      - SSD: 20GB BTRFS for setting (this way even if I'll need to format OMV I will keep them)
      - SSD: 12GB BTRFS swap for hibernation
      - HDD 4TB1: BTRFS this will the main 4TB
      - HDD 4TB2: BTRFS every night at 2am a job will sync from 4TB1 to 4TB2 so that 4TB2 is an exact copy of 4TB1
      - HDD 6TB1 and 2 exactly like the 4TB

      What do you think? Do you have any suggestion? Do you think that the sync job is a bad idea compared to RAID1 or ZFS Mirror?
      Intel G4400 - Asrock H170M Pro4S - 8GB ram - 2x4TB WD RED in RAID1 - ZFS Mirror 2x6TB Seagate Ironwolf
      OMV 4.1.4 - Kernel 4.14 backport 3 - omvextrasorg 4.1.2
    • From what I have read so far, there will be no way to upgrade an OMV 4 to an OMV 5, so, regardless of what filesystem you choose for the OS drive, you'll have to do a fresh install from scratch. So picking BTFS for OMV 4 just for this reason gets you nothing. Also, I have a very hard time believing that EXT4 will be completely abandoned. If I can't use my existing on OMV 4 data array of 32TB with OMV 5, I'll stick with OMV 4.

      You can't use the standard OMV installation iso images to partition the system disk into more than one partition. So to split your 60GB SSD into 20, 20, and 12GB partitions, you will first have to install some version of Debian Stretch and then install OMV into that.

      Not sure what you mean by using the second 20GB partition "for setting (this way even if I'll need to format OMV I will keep them)." I am not aware of any working strategy that allows you to start again with a fresh install of OMV and import "things" into it to restore it back the way it was. The only way I know how to do this is to make a dd type image of the entire OMV disk and restore that image to bare metal when needed. I currently make an automated on cron daily live dd image of my 16GB SSD OMV disk, keeping the most recent five daily images. This has been repeatedly verified to restore and is re-verified to work properly every sixty days for the last three years. This "costs" me only 80GB of disk space to do.

      Your decision to make one for one identical daily copies of your data disks is expensive. Also, you may want to be very sure that you can actually finish copying 10TB of data before it's time to make the next daily copy. And of course if you add or remove any data to the disks while they are being copied, those changes might not make it into the copies. I think you can do better with SnapRAID running one parity disk and syncing and scrubbing on a reasonable schedule. I SnapRAID diff, sync, and scrub every other day. The problem with RAID1 or other type of mirror is that if you make an unintentional deletion, it's more or less instantly removed from the mirror too. If the purpose for this was to be backup, it isn't.
      OMV 4.x - ASRock Rack C2550D4I - 16GB ECC - Silverstone DS380
    • Thanks for the help! I will answer to everything, hope to clear what my intentions are :)

      gderf wrote:

      From what I have read so far, there will be no way to upgrade an OMV 4 to an OMV 5, so, regardless of what filesystem you choose for the OS drive, you'll have to do a fresh install from scratch. So picking BTFS for OMV 4 just for this reason gets you nothing. Also, I have a very hard time believing that EXT4 will be completely abandoned. If I can't use my existing on OMV 4 data array of 32TB with OMV 5, I'll stick with OMV 4.
      github.com/openmediavault/openmediavault/issues/101
      I too agree that not support ext4 is a terrible decision BTW


      You can't use the standard OMV installation iso images to partition the system disk into more than one partition. So to split your 60GB SSD into 20, 20, and 12GB partitions, you will first have to install some version of Debian Stretch and then install OMV into that.
      I will just use gparted live after installing OMV


      Not sure what you mean by using the second 20GB partition "for setting
      (this way even if I'll need to format OMV I will keep them)." I am not
      aware of any working strategy that allows you to start again with a
      fresh install of OMV and import "things" into it to restore it back the
      way it was. The only way I know how to do this is to make a dd type
      image of the entire OMV disk and restore that image to bare metal when
      needed. I currently make an automated on cron daily live dd image of my
      16GB SSD OMV disk, keeping the most recent five daily images. This has
      been repeatedly verified to restore and is re-verified to work properly
      every sixty days for the last three years. This "costs" me only 80GB of
      disk space to do.
      I was talking about the application settings (plex, sonarr, deluge,ecc.)


      Your decision to make one for one identical daily copies of your data disks is expensive. Also, you may want to be very sure that you can actually finish copying 10TB of data before it's time to make the next daily copy. And of course if you add or remove any data to the disks while they are being copied, those changes might not make it into the copies. I think you can do better with SnapRAID running one parity disk and syncing and scrubbing on a reasonable schedule. I SnapRAID diff, sync, and scrub every other day. The problem with RAID1 or other type of mirror is that if you make an unintentional deletion, it's more or less instantly removed from the mirror too. If the purpose for this was to be backup, it isn't.
      I'm 100% that I won't have problem with the time of the copy, and for the hour: I will do it at 2 or 3am when I'll be in the dream world, so there is no risk I'll change anything during that time.
      Still I'm curios about snapRAID. How does it works? Is the parity disk and the files on it readble if I connect it to an other computer?
      I don't need a backup, as I said I have an external hdd for that. What I need is an easy way to save my data if an hdd die out of blue. Since I don't want a backup, but still the risk of deleting data by accident it exist, I was thinking about a daily mirror with an rsync job that run at night.
      Intel G4400 - Asrock H170M Pro4S - 8GB ram - 2x4TB WD RED in RAID1 - ZFS Mirror 2x6TB Seagate Ironwolf
      OMV 4.1.4 - Kernel 4.14 backport 3 - omvextrasorg 4.1.2

      The post was edited 2 times, last by Blabla ().

    • Plex doesn't have what I can describe as "portable" application settings. They are contained in with the metadatbase directory, and mine is well over 100GB and it grows about 5GB per day. The others have configuration directories that are portable, but if you are running these apps in dockers, you need the docker container data and OMV's docker plugin configuration (not portable) to go with them if you think restoring this stuff will be easy. I just don't see anything workable with what you are going to do with that other 20GB partition. Also, if the disk fails, it's all lost anyway, regardless of how many partitions are on it.

      It may be remotely possible to take a SnapRAID parity disk to another computer, but you will need more than that to be able to do anything with it. You'll need other related files, and it still might not work. The parity information is stored in a single large file - it's as large as your largest disk being protected by SnapRAID, in your case it could be a single nearly 6TB file. The way SnapRAID is used to restore lost data is to bring a new disk into the same machine and restore the data there. You can google snapraid to find its home page for all you will need to know about it.

      A SnapRAID setup with a single parity drive will protect you from the complete total loss of any one of your disks that have been synced. If you want to protect yourself from two simultaneous total disk failures, you will need two parity disks, and so on. I added another parity drive for a total of two when I added my fifth data drive.

      SnapRAID will protect you from data deletion, but only if the array was synced since that data was added.
      OMV 4.x - ASRock Rack C2550D4I - 16GB ECC - Silverstone DS380
    • Thanks a lot again :D
      Plex doesn't have what I can describe as "portable" application settings. They are contained in with the metadatbase directory, and mine is well over 100GB and it grows about 5GB per day. The others have configuration directories that are portable, but if you are running these apps in dockers, you need the docker container data and OMV's docker plugin configuration (not portable) to go with them if you think restoring this stuff will be easy. I just don't see anything workable with what you are going to do with that other 20GB partition. Also, if the disk fails, it's all lost anyway, regardless of how many partitions are on it.
      So if I install Plex with docker, delete the plex docker, than install it again and point it to the settings folder of the preview installation it won't work? I'm pretty sure that the metadabase directory is portable, and this one is actually important to me.
      Of course if the SSD will fail I'll lose everything,but in that I'll go with it. Maybe I'll do a backup of the SSD once in a while.



      It may be remotely possible to take a SnapRAID parity disk to another
      computer, but you will need more than that to be able to do anything
      with it. You'll need other related files, and it still might not work.
      The parity information is stored in a single large file - it's as large
      as your largest disk being protected by SnapRAID, in your case it could
      be a single nearly 6TB file. The way SnapRAID is used to restore lost
      data is to bring a new disk into the same machine and restore the data
      there. You can google snapraid to find its home page for all you will
      need to know about it.


      A SnapRAID setup with a single parity drive will protect you from the
      complete total loss of any one of your disks that have been synced. If
      you want to protect yourself from two simultaneous total disk failures,
      you will need two parity disks, and so on. I added another parity drive
      for a total of two when I added my fifth data drive.


      SnapRAID will protect you from data deletion, but only if the array was synced since that data was added.
      I'll check tomorrow, I need to understand if snapRAID is fine for me or if I'll go with an rsync plug-in.
      Intel G4400 - Asrock H170M Pro4S - 8GB ram - 2x4TB WD RED in RAID1 - ZFS Mirror 2x6TB Seagate Ironwolf
      OMV 4.1.4 - Kernel 4.14 backport 3 - omvextrasorg 4.1.2
    • Sorry, but I neglected to mention that taking a SnapRAID parity file to another machine and being able to use it to restore lost data would also mean you have to take all the remaining other disks that are part of the array too. That is to say that the parity file alone is not enough to restore a lost disk. The other remaining disks in the array are required to be present also if a restore is to be attempted.
      OMV 4.x - ASRock Rack C2550D4I - 16GB ECC - Silverstone DS380
    • Blabla wrote:

      Thanks a lot again :D
      So if I install Plex with docker, delete the plex docker, than install it again and point it to the settings folder of the preview installation it won't work? I'm pretty sure that the metadabase directory is portable, and this one is actually important to me.Of course if the SSD will fail I'll lose everything,but in that I'll go with it. Maybe I'll do a backup of the SSD once in a while.


      I'll check tomorrow, I need to understand if snapRAID is fine for me or if I'll go with an rsync plug-in.
      Yes, what you say will work, but only because you are going to reinstall the docker, and reconfigure the container. This is not the same thing as merely copying some files.

      Generally speaking rsync will be much more expensive than SnapRAID. By expensive, I mean the amount of disk space needed.

      My Plex metadatbase directory is not backed up or protected by SnapRAID. Its structure is not suitable for SnapRAID, and it can be regenerated if lost - there is no non-replaceable data in it.
      OMV 4.x - ASRock Rack C2550D4I - 16GB ECC - Silverstone DS380
    • gderf wrote:

      Sorry, but I neglected to mention that taking a SnapRAID parity file to another machine and being able to use it to restore lost data would also mean you have to take all the remaining other disks that are part of the array too. That is to say that the parity file alone is not enough to restore a lost disk. The other remaining disks in the array are required to be present also if a restore is to be attempted.
      That was I was looking for. I know that rsync is a lot more expansive but I'm fine with that. I look for something easy to set up and to use. I'll test snapRAID but I think that what I need is rsync even if it's a lot more expansive :/

      Do you have any other suggestion other that snapraid VS rsync? :)
      Intel G4400 - Asrock H170M Pro4S - 8GB ram - 2x4TB WD RED in RAID1 - ZFS Mirror 2x6TB Seagate Ironwolf
      OMV 4.1.4 - Kernel 4.14 backport 3 - omvextrasorg 4.1.2