Newer EC2 instances (r5b, r6i, m6i, c6i) use physical_block_size = 4096 for EBS volumes, whereas older instances (r5, m5, c5, i3en) use physical_block_size = 512. Below is a summary on using different sector size values in a cluster:
- It is safe to have mixed physical block sizes in the same ASM disk group.
- It is safe to change disk's physical block size (by changing EC2 instance type).
- secsize in disk's ASM metadata is set to match physical block size when creating an ASM disk group or adding a disk to an existing disk group.
- secsize in disk's ASM metadata is automatically updated both ways (512 <-> 4096) to match physical block size when offline disks in a mounted disk group are onlined.
- secsize in disk's ASM metadata is not updated from 512 to 4096 when mounting a disk group. But the mount will succeed.
- secsize in disk's ASM metadata is NOT updated from 4096 to 512 when mounting a disk group. Such mount attempt will FAIL.
- To switch from an instance with 4096 physical block size (r5b, r6i, m6i, c6i) to an instance with 512 physical block size (r5, m5, c5), it is mandatory to resize one node at a time while keeping the disk group mounted on at least one database node. Resizing instances while the cluster is stopped (disk groups not mounted) will result in failure to mount the disk groups after the resize.
Keyword: EC2 instance upgrade