On 2/15/26 23:57, Stuart Longland VK4MSL via Users wrote:
Hi Ján, apologies for the delay. Things got side-tracked my end.
On 9/2/26 17:15, Ján Tomko wrote:
To narrow it down, you can tell libvirt to write debug logs and share the part around starting the pool: https://libvirt.org/kbase/debuglogs.html
At a quick glance, I have not found any obvious place with a missing error message.
I ran:
virt-admin -c virtlogd:///system daemon-log-outputs "3:stderr 1:file:/ tmp/libvirtd-debug.log" virt-admin -c virtlogd:///system daemon-log-filters "1:storage" virt-admin -c virtlogd:///system daemon-timeout --timeout 0
and got back a 1.8GB log file in the space of 11 minutes.
The bit that seems to be relevant is this bit.
Indeed it is.
2026-02-06 20:40:30.941+0000: 3592: debug : virStorageBackendRBDRefreshPool:694 : Utilization of RBD pool one: (kb: 15628107776 kb_avail: 6502515640 num_bytes: 7893209706150)
<snip/>
2026-02-06 20:40:32.148+0000: 3592: debug : volStorageBackendRBDRefreshVolInfo:570 : Refreshed RBD image one/sjl-router-obsd76-vda (capacity: 34359738368 allocation: 34359738368 obj_size: 4194304 num_objs: 8192)
<snip/>
2026-02-06 20:40:32.261+0000: 3592: debug : volStorageBackendRBDRefreshVolInfo:570 : Refreshed RBD image one/sjl-router-obsd76-vda (capacity: 34359738368 allocation: 34359738368 obj_size: 4194304 num_objs: 8192)
For some reason, this volume is examined twice. And that makes virStoragePoolObjAddVol() fail (as we don't allow two volumes with the same name in one pool) which in turn makes virStorageBackendRBDRefreshPool() and subsequently pool start fail. I'll post o patch that adds appropriate error message. BUT, it will only make libvirt report an error. What we need to debug is, why libvirt got the same image twice.
Best I can see from that, it was *successfully* listing the RBD entries in the storage pool, then it gives up and declares it a failure with "no error".
And you were right :-) Michal