home :: FreeBSD :: newNAS2018

Setting up my new home NAS with FreeBSD 11.2

So, I've got my new HPE Proliant Microserver Gen10 last summer, but I hadn't had time to set up it yet.

I am setting it up with FreeBSD. (It's a sobering thought that FreeBSD is now 25 years old: I've been running it since the very first version, coming from 386BSD and Microport System V/AT before that.)

I've put the 4 Seagate drives in the bays and inserted a USB key with FreeBSD 11.2 on it in the internal USB port. As reported in the previous blog entry, I had to add a "/boot/loader.conf" file with

hw.pci.realloc_bars="1"

in it.

Then, I needed to:

• Install packages with pkg: perl5, zip/unzip, zile, mutt, rsync, sudo, samba, and smartmontools

• Copy the user accounts from my old proliant server onto the new datasink server. I haven't decided yet if I will shut down the old server or if I will just let it run on my network.

• Install the 4 Seagate drives under ZFS... more on this below.

• Copy the files from my old proliant server to the new one.

• Activate NFS on the new server. With ZFS, I don't need to start the usual NFS server processes, as ZFS supports NFS sharing directly (see below).

• Move the large external USB drive that I use for weekly rsync backups onto the new machine.

• Configure regular backups, monitoring, etc. on the new server.

• Exchange the names proliant and datasink on the two machines. Verify that the Windows machines and the Macs on my home network can access the new server transparently.

• And that should be it.

Setting up the 4 Seagate disks with ZFS

Here is how I set up the four 6TB disks on that new server:

1. Decided to use the 4 drives in a RAIS-5 configuration, which allows me to recover from a single drive failure. If the Microserver had had 5 drives, I would have used a RAID-6 configuration, but with 4 drives, it seemed like an exageration.

2. Decided to create a single large ZFS pool with all of the drives in it.

3. Also decided to limit the number of ZFS filesystems I create in my ZFS pool. For the limited usage I make of my little home server, I haven't found much use over the years for splitting the disk space pool into multiple logical filesystems, apart from sharing via NFS, so this time, I'm going with just a few filesystems: a large one called /data, a smaller one for user files called '/homes' that will be exported via NFS, and the usual /var and /usr.

4. However, I've been running a ZFS snapshot script on my machine for the past few years (as explained here), and I've found it quite useful. So I decided to continue with them. So I must account for enough freespace in the ZFS pool for all of the data from the snapshots.

5. For other uses, I'll create symbolic links into the /data zpool for /backups and others.

6. Note: I have used the /dev/diskid/* devices to set up my ZFS pool, rather than the /dev/ada* devices. If I get a drive failure or take a drive out, I find the renumbering of ada devices to be a scary proposition. Similarly, when I use external USB drives, I give them GPT names and use these rather than /dev/da* which can change when you unplug/replug the USB connectors.

So, here are the commands that I've run to initialize the disk space on my new server, conserved here for future reference:

• Check out which disks we have available:

  ls /dev/diskid

• Create the ZFS pool (the ashift option specifies the drives have 4KB blocks)

  zpool create -o ashift=12 -m /zfspool zfspool raidz diskid/DISK-ZAD4QZAT diskid/DISK-ZAD4RS61 diskid/DISK-ZAD4V4N3 diskid/DISK-ZAD4V4N9
  (question: why have a mountpoint for a ZFS pool?)

• Create the filesystems in my ZFS pool:

  zfs create zfspool/data mountpoint=/data
  zfs create zfspool/var mountpoint=legacy
  zfs create zfspool/home mountpoint=legacy
  zfs create zfspool/usr mountpoint=legacy
  zfs create zfspool/tmp mountpoint=legacy

• Copy the existing stuff onto the new partitions and set the real mountpoints:

  foreach i ( var home usr tmp )
  mount -t zfs zfspool/$i /mnt
  ( cd /$i ; sudo rsync -aHv . /mnt/ )
  umount /mnt
  zfs set mountpoint=/$i zfspool/$i
  end

• Share the /home dataset:

  zfs set sharenfs=on zfspool/home
  (check that this works... I am not sure)

/FreeBSD | Posted at 07:39 | permanent link