lvm volume seemingly present after vm delete

[ioo]

Hi!

I would like to ask about one phenomena which does not excatly break anything but is at times unconvinient. I have encountered it several times so i dont think it is specific to my particular case.

block device is connected over iscsi to multiple pve cluster nodes and on this block device is lvm (virtual computer block devices are actually node lvm volumes). In addition in my case at pve host there is multipath between iscsi and lvm. I think it is very usual thing, i have used it many years, works good.

At times i found system in such i state that there are present entries in 'dmsetup info -c' which obviously belonged to virtual machines but those machies are already deleted

# dmsetup info -c
Name                          Maj Min Stat Open Targ Event  UUID                                                               
mp_eternus_1                  253   6 L--w    9    1      0 mpath-3600000e00d300000003008f000030000                             
vg_eternus_1-vm--325--disk--2 253 152 L--w    0    2      0 LVM-uv1bq9ve2LZrt0KAwt26X3AmQBqfbfuzZjjXfmG1Jvu0ZspRJdWRMbkckuzrkRIi
vg_eternus_1-vm--398--disk--0 253 252 L--w    0    3      0 LVM-uv1bq9ve2LZrt0KAwt26X3AmQBqfbfuzYfetm2QcZf9igQyeeXbRPzFxe3MywTeZ
vg_eternus_1-vm--677--disk--0 253 333 L--w    0    1      0 LVM-uv1bq9ve2LZrt0KAwt26X3AmQBqfbfuzJRDi3aJoN8N52yelVKwVkVUmShM8Htgk
vg_eternus_1-vm--677--disk--1 253 334 L--w    0    1      0 LVM-uv1bq9ve2LZrt0KAwt26X3AmQBqfbfuzsdBQWTCaa0lUFN3R2xUISyznBa2zAQ6o

In lvdisplay output no such entries exist. Problem appears when i would like to disconnect for some reason (like maintenance) this node from iscsi resource. Generally i say

# vgchange -a n
# multipath -f mp_eternus_1
# iscsiadm -m node -T iqn.2022-07.com.fujitsu:storage-system.eternus-dxm:vol-hdd-proxmox:0010 -p 10.28.4.21:3260 -u
# iscsiadm -m node -T iqn.2022-07.com.fujitsu:storage-system.eternus-dxm:vol-hdd-proxmox:0110 -p 10.28.8.21:3260 -u

But in the event of resources being present in dmsetup-info-c output i get

# multipath -f mp_eternus_1
Oct 31 17:49:43 | mp_eternus_1: map in use

I imagine there are forceful ways to solve it and maybe it is appropriate in this case, for example in the spirit of what good people describe here, https://elkano.org/blog/removing-multipath-device/

I would be thankful if you could help me forward with this.


Best regards,

Imre

 

[fireon]

The behavior sounds like "saferemove_throughput". In the WebUI you can activate this feature directly under "Datacenter -> Storage -> your LVM entry". There this is called "Wipe throughput".