如何替换含数据的预测错误坏磁盘（含5PV的VG场景）

Got it, let's break down exactly how to replace that problematic PV (throwing predictive failure alerts) in your 5-PV volume group. This is a common scenario, and as long as you follow the steps carefully, you can pull this off without data loss.

Step 1: Identify the Failing PV & Verify Your Setup #

First, you need to pinpoint the bad disk and confirm your VG/LV health:

• Run pvs to list all physical volumes—look for flags like failed or cross-reference with system logs (check dmesg or smartctl output) to find the disk with predictive errors.
• Use vgs to check your volume group's free space: you’ll need enough free physical extents (PE) on the remaining 4 healthy PVs to hold all data from the failing disk.
• Run lvdisplay to see how your logical volumes are spread across the PVs—this helps you confirm there’s no dependency tied exclusively to the bad disk.

Step 2: Migrate Data Off the Failing PV #

Once you’ve confirmed the failing PV (let’s say it’s /dev/sdX), use pvmove to transfer all its data to the other healthy PVs:

pvmove /dev/sdX

• This command automatically distributes the data to free space on the remaining PVs. If you want to target a specific healthy PV (e.g., /dev/sdZ), run pvmove /dev/sdX /dev/sdZ instead.
• If the migration gets interrupted, you can resume it with pvmove --resume or abort cleanly with pvmove --abort.

Step 3: Remove the Failing PV from the VG #

After the data migration finishes successfully, detach the bad PV from your volume group (replace myvg with your actual VG name):

vgreduce myvg /dev/sdX

• Double-check with pvs to make sure /dev/sdX is no longer linked to the VG.

Step 4: Physically Swap the Disk #

Power down your system (or use hot-swap if your hardware supports it), pull out the failing disk, and insert the new replacement drive.

Step 5: Initialize the New Disk as a PV #

Once the system recognizes the new disk (use lsblk or fdisk -l to get its device name, e.g., /dev/sdY), set it up as a physical volume:

pvcreate /dev/sdY

• If the new disk has old partitions, wipe them first with dd if=/dev/zero of=/dev/sdY bs=1M count=10 (warning: this erases all data on the disk—be sure you’re targeting the right device!).

Step 6: Add the New PV to the VG #

Extend your volume group to include the fresh PV:

vgextend myvg /dev/sdY

Step 7: Verify Everything’s Working #

Wrap up by confirming the setup is healthy:

• Run pvs to see the new PV listed under your VG with available space.
• Use vgs to check that the VG’s total size matches the sum of all 5 PVs (now with the new disk replacing the old one).
• For file systems on your logical volumes, run a integrity check: e2fsck /dev/myvg/mylv for ext4, or xfs_check /dev/myvg/mylv for XFS.

Quick Pro Tips #

• Backup First: Even though pvmove is reliable, always back up critical data before messing with disk replacements—better safe than sorry.
• Low Traffic Time: Run the pvmove process during a quiet period; it’s resource-heavy and can slow down your system if done during peak usage.
• Monitor Progress: Keep an eye on the migration with pvs or vgs—large datasets can take a while to transfer.

内容的提问来源于stack exchange，提问作者Franklin Kabiladhas