This is my first manually created ZFS pool experience. Before I get into the details, let me say this is on FreeBSD 10.1 , so if you are trying to use this as a guide it may not work on other versions. As well I am not responsible to anything that may occur if you try any of these commands, etc. Use this information at your own risk.
So I figured my drives were 4k sector, but I thought I would double check to make sure. I ran "camcontrol identify da0", and looked for the line "sector size logical 512, physical 4096, offset 0". As you can see the physical sector is 4k. So vfs.zfs.min_auto_ashift (a kernel parameter) by default is set to 9 which is 512, so I changed it to 12 which is 4k.
Now I created all my GPT (GUID Partition Table ) for each disk. For those just catching my posts for the first time I have six 3TB Western Digital Red drives. This is "gpart create -s gpt
So looking at my device names, they were generic da0-5. I thought if my ZFS pool has an issue in the future how on earth do I know which drive to replace. So I tested this out a little, I unplugged the top drive in my case, then I checked dmesg. Guess what, the drive I would like to be drive 1 is da5. So my schema for labeling will be simple, since all of my disks are internal for now. I'm going to label them internal disk # partition #. So in my case it was "gpart add -a 4k -l id1p1 -t freebsd-zfs da5". If you want to view the labels look at "glabel status". I will follow the same format for all my drives. This way when I do have an issue, I'm not trying to figure out which disk is which. I figure it's better to find the information while the device is actually working and commands will run against it. If it does fail later on and I have a degraded pool, unplugging a good drive could lead to more issues. For reference of how crazy it can be da3 was disk 2, da2 was 3, da4 was 4, da1 was 5, and da0 was 6. And yes I know that re-cabling could help with this, but storing the labels on the drives has several advantages. For starters I can move them around all I like and the labels will be there. Second mine could be matched up but only because they started at da0. If I used other controllers, or hooked internal connections up to say ESATA, or some other external device, I may wish to label it in more detail, like enclosure 1 drive 1 or e1d1p1.
Anyways enough about labels. Once you have them all labeled. Now we have our labels and partitions. It's time to make the zpool now. I have decided on RAIDZ2. Here goes nothing, "zpool create storage raidz2 gpt/id1p1 gpt/id2p1 gpt/id3p1 gpt/id4p1 gpt/id5p1 gpt/id6p1". It worked, no warnings, no errors. With the overhead, hard drive space rounding and my 4k aligning of the drives, total usable is about 10.5TB. Not too bad.
zpool list NAME SIZE ALLOC FREE FRAG EXPANDSZ CAP DEDUP HEALTH ALTROOT storage 16.2T 756K 16.2T 0% - 0% 1.00x ONLINE - zpool status NAME STATE READ WRITE CKSUM storage ONLINE 0 0 0 raidz2-0 ONLINE 0 0 0 gpt/id1p1 ONLINE 0 0 0 gpt/id2p1 ONLINE 0 0 0 gpt/id3p1 ONLINE 0 0 0 gpt/id4p1 ONLINE 0 0 0 gpt/id5p1 ONLINE 0 0 0 gpt/id6p1 ONLINE 0 0 0 zfs list NAME USED AVAIL REFER MOUNTPOINT storage 480K 10.5T 192K /storage
Now that I have my space, I can start working on defining some datasets and jails, etc. But that will be another day.