In a prior incarnation of my “RAID” system, I was using 1 partition on one drive plus 4 partitions on another drive to set up a RAID-Z2 system.
My thinking was that splitting up by partition at least protects me against sector failure (if not whole drive failure). The 4 partitions were 250 GB each on a 1TB Hitachi/SimpleTech drive.
I should’ve anticipated it, but don’t you know, the 1 TB drive failed–after about 1 week. So, my data would’ve been lost. (Except I kept a copy on another computer; I’m not crazy.)
I knew I’d need at least 3 hard drives to prevent this from happening again. So, I sent the Hitachi drive for warranty and ordered a Fantom 1 TB hard drive. (BTW: I really like the Fantom; it looks slick, is constructed well, and runs rather cool.)
In the meantime, I spotted a 500 GB Fantom for a good price (not too enthused about the mail-in-rebates, though), so I got that as well. Once I had the two Fantoms in hand, I could set up a RAIDZ (2 data + 1 parity) configuration with what I already had: a 250 GB Cavalry hard drive (from before), a 500 GB Fantom, and a 1 TB Fantom. Setting it up this way meant that I’d only be making use of 750 GB of space (out of the 1.75 TB that all the drives have together). 250 GB of that 750 is used for parity, so I really only get 500GB of usable space. Such is the price for reliability.
A couple days ago, the replacement 1 TB Hitachi arrived. So, I decided to take out the 250 GB hard drive and replace that slice with the 1 TB Hitachi. I did a:
sudo zpool offline tank da1s1
I then powered down and unplugged the 250GB Cavalry drive. Seeing as how FreeBSD can be quite flaky in disk numbering, I made sure to keep track of which USB port it was plugged into.
Incidently, I had a Kill-A-Watt on the whole setup. Before unplugging the Cavalry, the Kill-A-Watt estimated $36.33 per year and 33 Watts usage. After unplugging, it said 22 Watts usage. Wow! That explains why the Cavalry runs burning hot to the touch (makes a good foot warmer in winter).
After plugging in the Hitachi drive (I inadvertently powered down the machine in the process), I did the following to re-silver:
sudo zpool replace tank da1s1 da1s1
Right now, it’s sitting at the following:
> zpool status
pool: tank
state: DEGRADED
status: One or more devices is currently being resilvered. The pool will
continue to function, possibly in a degraded state.
action: Wait for the resilver to complete.
scrub: resilver in progress for 0h12m, 11.90% done, 1h35m to go
config:
NAME STATE READ WRITE CKSUM
tank DEGRADED 0 0 0
raidz1 DEGRADED 0 0 0
da0s1 ONLINE 0 0 0 4.89M resilvered
replacing DEGRADED 0 0 0
da1s1/old OFFLINE 0 7.68K 0
da1s1 ONLINE 0 0 0 7.23G resilvered
da2s1 ONLINE 0 0 0 4.88M resilvered
errors: No known data errors
Pretty slick! I’ll now have 1 TB of storage (out of the 2.5 TB total hard drive space).
I know it seems crazy, but I think I’ll continue to keep the slices incongruent. For example, when I need more space, I’ll replace the 500 GB drive with a 2 TB drive. I’ll immediately have 2 TB available (out of a total of 4 TB hard drive space, 3 TB are being used). One of the mistakes I made presently is that I have two drives of the same size. So, at some point, I’m going to replace one of them but not get any additional space from it. More likely, 2 TB of space will suffice for many years to come.
By the way, the Kill-A-Watt now shows 30 Watts usage. Of course, that’s while the drives are active (re-silvering). Hopefully, they’ll lower their power when the re-silvering is finished.
Post a Comment