[linux-lvm] question about vg underlying pv devs order

[heming.zhao at suse.com]

Hello list,

one our customer reported a issue. After executing lvextend, the vg underlying pv devs changed.

I do a summary from my watching.

first, pv devies in /etc/lvm/backup/systemvg & /etc/lvm/archive/systemvg-xxx are correct.

but using pvdissect (url: https://wiki.syslinux.org/wiki/index.php?title=Development/LVM_support/pvdissect), the script dumped show pv related device had been changed after executing lvextend command.

the part output of pvdissect like:
```
0x00006000 (metadata.value):
systemvg {
     id = "id-xxx"
     seqno = 11
     format = "lvm2"
     status = ["RESIZEABLE", "READ", "WRITE"]
     flags = []
     extent_size = 8192
     max_lv = 0
     max_pv = 0
     metadata_copies = 0

     physical_volumes {

         pv0 {
             id = "pv0-uuid-xxxx"  <=== it didn't change
             device = "/dev/sdb"   <=== it changed to /dev/sda

             status = ["ALLOCATABLE"]
             flags = []
             dev_size = 4806528
             pe_start = 2048
             pe_count = 586
         }

         pv1 {
             id = "pv1-uuid-xxxx"   <=== it didn't change
             device = "/dev/sdc"    <=== it changed to /dev/sdb

             status = ["ALLOCATABLE"]
             flags = []
             dev_size = 4806528
             pe_start = 2048
             pe_count = 586
         }

         pv2 {
             id = "pv2-uuid-xxxx"    <=== it didn't change
             device = "/dev/sdd"     <=== it changed to /dev/sdc

             status = ["ALLOCATABLE"]
             flags = []
             dev_size = 4806528
             pe_start = 2048
             pe_count = 586
         }

     ... ... # other parts are correct
```

(the systemvg consists of many LVs, ie: usrlv, varlv, optlv, homelv.)
  

my question:
the devs order changed, but after system reboot, everything work fine. why?

I not very familiar with lvm code.
 From my watching, it looks lvm uses pv uuid not device to work, is it right?

Thanks.