A few days ago I was down in the server room, and I noticed an orange light next to the top disk in Drobo2 (my second Drobo box). I brought up the Drobo dashboard, and it said that my disks were OK, but that the array was getting full, and I should install a larger capacity disk in the slot next to the orange light.
That was kind of a surprise. I keep thinking of the Drobos as RAID boxes with a few bells and whistles, but this wasn’t anything I’d expect a RAID box to do. I’m sure the idea of incremental capacity increases is a good thing in some applications, but it doesn’t help me, since my backup scheme depends on my being able to take the whole array to the safe deposit box.
I ordered five 2 TB disks, four for the array and a spare. This time there was no question that I wanted the Western Digital green disks, since, no matter how much their performance is compromised in the search for low power and long life, their speed is not going to be tested when they’re in a storage device that connects to the host via USB.
The disks arrived promptly. I disabled all the Vice Versa scripts that referenced Drobo2. I wrote down the top-level directory names in Drobo2. I tried to put the array into standby, but I kept getting an error message indicating that the drive was in use. Finally, I powered down the server, popped out the old drives, installed the new ones, and rebooted. While I waited I tucked the old drives into conductive plastic shock cases for the trip to the bank.
With the computer up, I set up an 8GB partition. The Drobo dashboard warned me that that could increase boot time, but it sure makes managing the backups easier. Working from my list, I recreated all the top-level directories. Then I enabled the scripts one as a time, running them manually and making sure they completed before enabling the next one. I did this to keep two scripts from contending for the heads, which would slow down the restore and create more fragmenting.
It took a day and a half to restore 2.5 GB of data from the Serial SCSI drives on the server to the USB Drobo. Note that the throughput is about twice as fast as the network restore from the same Serial SCSI drives to a single SATA disk that I did last week. Why gigabit Ethernet provides throughputs of a bit over 100 megabits/sec when the computers are as fast as they are is beyond me. I’ve seen sustained throughput as high as 300 Mb/s, but I can’t count on it.
All in all, it was fairly painless. After my earlier struggles with the Drobos, I’m becoming a believer.
Brad Meyer says
Did you know that you could have inserted the new 2TB drives while the system was up and running? The orange light indicates that you should replace that drive with a larger one and you can simply unplug the existing drive and insert the new drive while the Drobo is up and running. That’s the beauty of BeyondRAID technology. The system will automatically protect itself when it sees the old drive removed and then automatically accept the new larger drive once it is inserted. It will then restripe the data as necessary with the new drive capacity and return the system to complete data safety with all green lights. Once you see all green lights again you can hot-replace the next drive with a larger one. Very simple.
Yes, I did know that, and I referred to it in my original post when I said: “I’m sure the idea of incremental capacity increases is a good thing in some applications, but it doesn’t help me, since my backup scheme depends on my being able to take the whole array to the safe deposit box.”
Maybe you don’t understand that my off-site backup scheme requires that I take a set of drives out of the Drobo and put them in a safe deposit box. If I replaced the drives one at a time, the drives that I removed would not be a set.
There’s another problem, I think. If I replaced the drives one at a time, wouldn’t my partition size be the same as it was with the older, smaller drives?