Creating Pool Storage: got Error: failed to notify peer 192.168.3.102:8443: Custom loop file locations are not supported

Incus
1

Hello forgive my noobness,
I have 2 hosts : “incus” and “cm-adrien” in cluster mode with some lxcs.
I tried to create a zfs storage pool typing in “cm-adrien”'s cli, doing:

[alain@ADRIEN:~]$ incus storage create --target incus pool1T zfs source=/dev/sdb1
Storage pool pool1T pending on member incus

[alain@ADRIEN:~]$ incus storage create --target cm-adrien pool1T zfs source=/dev/sdb1
Storage pool pool1T pending on member cm-adrien

[alain@ADRIEN:~]$ incus storage create pool1T zfs
Error: failed to notify peer 192.168.3.102:8443: Custom loop file locations are not supported

[alain@ADRIEN:~]$ sudo zpool list
NAME     SIZE  ALLOC   FREE  CKPOINT  EXPANDSZ   FRAG    CAP  DEDUP    HEALTH  ALTROOT
pool1T   912G   674K   912G        -         -     0%     0%  1.00x    ONLINE  -

[alain@ADRIEN:~]$ incus storage list
+---------+--------+-------------+---------+---------+
|  NAME   | DRIVER | DESCRIPTION | USED BY |  STATE  |
+---------+--------+-------------+---------+---------+
| default | btrfs  |             | 6       | CREATED |
+---------+--------+-------------+---------+---------+
| pool1T  | zfs    |             | 0       | ERRORED |
+---------+--------+-------------+---------+---------+

Why am I gettting this error? Thx.
Nota: 192.168.3.102 is “INCUS”'s IP. tcpport 8443 is allowed in that machine.

2

Sounds like the Incus server at 192.168.3.102 (not sure if that’s incus or cm-adrien) doesn’t have a disk at /dev/sdb1.

3

thx for your answer. You mean both hosts must have a separate disk plugged on? Or with one disk (in cm-adrien for my case) it is enough?
What I did is creating the pool in ONE disk only (cm-adrien : /dev/sdb1). I did lsklk to check on cm-adrien.

When creating storage pool, does the source depends on the target?

4

Yes, the source property is target specific, so the commands you ran specifically are telling Incus to expect a /dev/sdb1 on each server.

Storage pools must be consistent across the cluster, so all servers must be able to offer that pool to their instances, but you can mix what backs the pool, so on one server you can use a partition (source=/dev/sdb1) while on another you just have Incus create a basic loop disk (source= size=20GiB)

5

SOLVED
Thx Stéphane. I read the doc a little bit too fast, see:

incus storage create --target server1 data zfs source=/dev/vdb1
incus storage create --target server2 data zfs source=/dev/vdc1 <— not vdb1 !!
incus storage create --target server3 data zfs source=/dev/vdb1 size=10GiB

You may add your words : " source property is target specific". :slight_smile: