Support #40302
openrbd-mirror: how to ensure the data consistency in secondary cluster
0%
Description
When we are using rbd-mirror, we found a problem when we are going to start service on secondary image, where device-mapper reports data corruption.
I understand rbd-mirror can not know the transaction in upper layer application, but is there any suggestion to solve this problem?
What I can think about it is to add snapshot on primary periodly, then rbd-mirror will mirror the snapshot to secondary.
Yes, we could meet corruption in snapshot, but we can keep about 10 snapshots, we can start service if there is any one snapshot is consistent on device-mapper.
@Jason Borden, is that the correct way to use rbd-mirror? is there another way to solve it?
Thanx in advance!
Updated by Yang Dongsheng almost 5 years ago
Hi Jason,
After a second thought, I think that's not a good idea. We really should not wish an application-consistent with a crash-consistent technology (rbd-mirroring).
So we can close this issue now. Thanx
Yang Dongsheng wrote:
When we are using rbd-mirror, we found a problem when we are going to start service on secondary image, where device-mapper reports data corruption.
I understand rbd-mirror can not know the transaction in upper layer application, but is there any suggestion to solve this problem?
What I can think about it is to add snapshot on primary periodly, then rbd-mirror will mirror the snapshot to secondary.
Yes, we could meet corruption in snapshot, but we can keep about 10 snapshots, we can start service if there is any one snapshot is consistent on device-mapper.@Jason Borden, is that the correct way to use rbd-mirror? is there another way to solve it?
Thanx in advance!