I have a 3 node Proxmox cluster. The same nodes host a glusterfs volume setup as replicated 1 x 3 brick. Each brick on each host has 3 SATA SSDs managed by LVM and formatted as XFS.
On each node one of the SSD is wearing out faster than the other 2 SSD drives. On one host, it's above 80% and looking at the trend it might be a couple weeks or a month tops before wear level goes 100%. Hence, I need to prepare to replace each drive before wear level goes 100%.
I have not done this before. I am looking for suggestion on how to do this safely and quickly. So far I see 2 approaches.
1. At brick level. S for each host, remove the brick from the cluster. Replace the disk then add the brick back to the cluster. Repeat until each node has the worn out disk is replaced.
2. At the LVM level. Add the new disk to the volume group on each node. Migrate the data from the old disk to new one (pvmove). Then remove the old disk from the volume group (vgreduce/pvremove).
Which one is the better approach? Or is there a better one other than the 2 above I am thinking about?
On each node one of the SSD is wearing out faster than the other 2 SSD drives. On one host, it's above 80% and looking at the trend it might be a couple weeks or a month tops before wear level goes 100%. Hence, I need to prepare to replace each drive before wear level goes 100%.
I have not done this before. I am looking for suggestion on how to do this safely and quickly. So far I see 2 approaches.
1. At brick level. S for each host, remove the brick from the cluster. Replace the disk then add the brick back to the cluster. Repeat until each node has the worn out disk is replaced.
2. At the LVM level. Add the new disk to the volume group on each node. Migrate the data from the old disk to new one (pvmove). Then remove the old disk from the volume group (vgreduce/pvremove).
Which one is the better approach? Or is there a better one other than the 2 above I am thinking about?
