Growing disks: upgrading from 1 TB to 2 TB disks
We’ve got a fairly layered set of software that runs beneath our file systems: RAID1 followed by cryptsetup, followed by lvm. It works remarkably well.
However, when it’s time to do something like enlarge a partition or disk, it gets complicated.
We have a server (hubert) that was running two 1 TB disks in a RAID1. We typically create two partitions: one small boot partition and one large partition the uses the rest of the available disk (with 2 TB disks we’ve added a third partition for grub).
Last month I replaced one of the disks with a 2 TB disk, partitioned the new disk with similar a partition scheme (except that the “rest of the disk” partition was twice the size as the original one), and then added the new partitions to the RAID (mdadm /dev/md0 –add /dev/sda1; mdadm /dev/md1 –add /dev/sda2).
Yesterday, I replaced the 2nd 2 TB disks, and again added the new partitions.
The result was a fairly painless transition from 1 to 2 TB disks with minimal downtime.
The problem - though - is that we’re not using all the disk space available, since md1, the RAID array that uses the “rest of the disk” partition, is only about 1 TB when it should be just under 2 TB.
That problem was easy enough to solve by growing the RAID device with:
mdadm -G /dev/md1 -z max
It took a while, but eventually, running:
Reported the right sizes for md1.
After RAID, comes cryptsetup (/dev/md1 is the source device for md1_crypt). So, the next command was:
cryptsetup resize md1_crypt
In the output of:
I saw the change reflected in the dm-0 device.
Since md1_crypt is used as the physical volume for a volume group, the next step was to resize the physical volume:
This change was reflected with:
Now, the command:
Displayed a full 931 GB available that wasn’t there before.
Next up: resizing the logical volume:
lvresize --size 1700GB vg_hubert0/blanco_members
And, last but not least, was the file system:
See also dkg’s article on resizing crypt partitions.