View source for Zfs configurations: Storage Pool (FreeBSD)

= Documentation =
<blockquote>
{| class="wikitable"
|-
| create/grow mirrored pool || https://www.freebsd.org/doc/handbook/zfs-zpool.html
|-
|}
</blockquote><!-- documentation -->

= Tutorials =
<blockquote>
{| class="wikitable"
|-
| growing zpool || https://tomasz.korwel.net/2014/01/03/growing-zfs-pool/
|-
| replacing HDD || http://ryan.himmelwright.net/post/replace-zfs-mirror-drive/
|}
</blockquote><!-- Tutorials -->

= Creating & Naming Partitions =
<blockquote>

== Creating Partitions ==
You should create partitions that are slightly smaller than the total disk-size.
(by a few megabytes). This way you can account for small differences in avail bytes on future replacement drives.

<source lang="bash">
gpart show

# create GPT 
gpart destroy -F ada3
gpart create -s gpt ada3

# [6T drives]
    # ada1 WD-WXN1H842YH9K
    # ada2 WD-WXB1HB4L2K9C
    gpart add -s 5570G -a 4k -t freebsd-zfs ada1

# [8T drives]
    # ada1p1
    # ada2p1
    gpart add -s 7300G -a 4096 -t freebsd-zfs ada1

# [10T drives]
    # ada3p1 VCG7A2NN
    # ada4p1 VCG7J54N
    gpart add -s 9200G -a 4k -t freebsd-zfs ada3
</source>

== Disk Labels ==
As your storage array grows in size, it is handy to have something more informative 
than randomly assigned da1, da2, da3, ... for your disks. Create a naming/numbering scheme
based on the harddrive location in the box.<br>

Manually change the SATA connections, and reboot machine until the Disk-Order
is ascending (/dev/da0 at the top). You can then label your disks with:

{{ NOTE |
This type of disk labels does not persist. Look up permanent disk labels instead.
}}

<source lang="bash">
camcontrol identify /dev/ada2 | grep -i serial          ## displays same serial num as printed on physical disk

glabel create   s0d0-2-4HAx0H    /dev/ada0p2            ## create label (shelf0, disk0, partition2, last 6x digits of hdd serial (printed on disk))
glabel destroy /dev/label/s0d0-2-4HAx0H                 ## destroy label
</source>

== Mirrors and VDEVs ==
A mirror is a set of 2x, or 3x disks that all store the same contents, and each
disk can replicate the other if it fails. a VDEV is a raw device (file,disk,partition, etc).

<source lang="bash">
# create a mirror from regular zfs partition
# (take not of your disk-names, you'll need them)
zpool status

#            <pool-name>     <existing vdev>               <drive or partition to add>
zpool attach   zroot     diskid/DISK-WD-WX61DA4HAX0Hp3    /dev/label/s0d1-1-4BLJ4L
</source>

Creating a new mirror, and adding it to an existing pool
<source lang="bash">
zpool add  zroot  mirror    <disk_1>  <disk_2>
</source>
</blockquote><!-- Creating & Naming Partitions -->

= Replacing HDDs =
<blockquote>
== Configure zpool ==
<source lang="bash">
# zpool should consume extra space on avail partition
sudo zpool set autoexpand=on zroot  
</source>

== Identify HDDs ==
<source lang="bash">
# identify oldest HDDs to replace
# (try for all HDDs, in my case `ada${N}` )
smartctl -a /dev/ada1 | grep -i 'power_on_hours'
camcontrol identify /dev/ada1 | grep 'serial number'  # get serial num (on HDD label)
</source>

== Partition HDD ==
See above for creation instructions.

== Replace HDD ==
<source lang="bash">
# find GUID ('12345...') of prev HDD
zpool status

# swap HDDs within mirror
zpool replace ${zpool} ${guid} ada4p1

# wait for resilver
zpool status

# once everything works, remove original disk (NOTE: guid, not /dev!)
sudo zpool detach ${zpool} ${guid}
</source>
</blockquote><!-- Replacing HDDs -->