Created attachment 777803 [details] db dump, engine, vdsm logs Description of problem: Creating live snapshot of vm with two disks and then removing the vm leaves disks in system. These disks cannot be seen in "Disks" collection but only on storage domain (ie empty Disks tab, but Disks tab on Storage domain tab shows disks with no name). These disks cannot be removed and whole datacenter must be forcibly removed. Logical volumes of snapshots are left on storage as well: #lvs 59158d38-028e-4a40-9e40-0198021b1cae 5cf50403-521d-43d8-9318-6229013a1b60 -wi------ 1.00g 6c0b1839-ad3e-4eb6-9316-a11d21b3db5a 5cf50403-521d-43d8-9318-6229013a1b60 -wi------ 1.00g 7380017e-8202-48fc-ae8f-3f1f6c52eebb 5cf50403-521d-43d8-9318-6229013a1b60 -wi------ 1.00g e313df65-5e8a-452f-b1c4-89363e611917 5cf50403-521d-43d8-9318-6229013a1b60 -wi------ 1.00g But cleaned from dm: # lsblk NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT sda 8:0 0 232.9G 0 disk ├─sda1 8:1 0 250M 0 part /boot └─sda2 8:2 0 232.7G 0 part ├─vg_sf02-lv_root (dm-0) 253:0 0 69.5G 0 lvm / ├─vg_sf02-lv_swap (dm-1) 253:1 0 1G 0 lvm [SWAP] └─vg_sf02-lv_home (dm-2) 253:2 0 200M 0 lvm /home sdc 8:32 0 21G 0 disk └─36006048c0714062ba4b91f6a904bc07f (dm-4) 253:4 0 21G 0 mpath sdd 8:48 0 21G 0 disk └─36006048cc786b195adb3de2b6d613cbc (dm-8) 253:8 0 21G 0 mpath sdb 8:16 0 200G 0 disk └─36006048c78acaa6eac8ebc05bf73dee1 (dm-5) 253:5 0 200G 0 mpath ├─1b573d9d--6af4--4135--b8ec--6880e9d4d359-metadata (dm-15) 253:15 0 512M 0 lvm ├─1b573d9d--6af4--4135--b8ec--6880e9d4d359-leases (dm-16) 253:16 0 2G 0 lvm ├─1b573d9d--6af4--4135--b8ec--6880e9d4d359-ids (dm-17) 253:17 0 128M 0 lvm ├─1b573d9d--6af4--4135--b8ec--6880e9d4d359-inbox (dm-18) 253:18 0 128M 0 lvm ├─1b573d9d--6af4--4135--b8ec--6880e9d4d359-outbox (dm-19) 253:19 0 128M 0 lvm └─1b573d9d--6af4--4135--b8ec--6880e9d4d359-master (dm-20) 253:20 0 1G 0 lvm sde 8:64 0 21G 0 disk └─36006048c12952607a6254682d950135a (dm-6) 253:6 0 21G 0 mpath sdf 8:80 0 250G 0 disk └─36006048c14e9eb8f668dfc53ea5995ca (dm-7) 253:7 0 250G 0 mpath ├─5cf50403--521d--43d8--9318--6229013a1b60-metadata (dm-9) 253:9 0 512M 0 lvm ├─5cf50403--521d--43d8--9318--6229013a1b60-leases (dm-10) 253:10 0 2G 0 lvm ├─5cf50403--521d--43d8--9318--6229013a1b60-ids (dm-11) 253:11 0 128M 0 lvm ├─5cf50403--521d--43d8--9318--6229013a1b60-inbox (dm-12) 253:12 0 128M 0 lvm ├─5cf50403--521d--43d8--9318--6229013a1b60-outbox (dm-13) 253:13 0 128M 0 lvm └─5cf50403--521d--43d8--9318--6229013a1b60-master (dm-14) 253:14 0 1G 0 lvm /rhev/data-center/mnt/blockSD/5cf50403-521d-43d8-9318-622901 sdg 8:96 0 21G 0 disk └─36006048ce7ff8320bd02378f18cf9712 (dm-3) 253:3 0 21G 0 mpath # dmsetup ls 5cf50403--521d--43d8--9318--6229013a1b60-leases (253:10) 36006048c12952607a6254682d950135a (253:6) 36006048ce7ff8320bd02378f18cf9712 (253:3) 36006048c0714062ba4b91f6a904bc07f (253:4) 36006048c78acaa6eac8ebc05bf73dee1 (253:5) 1b573d9d--6af4--4135--b8ec--6880e9d4d359-master (253:20) 5cf50403--521d--43d8--9318--6229013a1b60-inbox (253:12) 1b573d9d--6af4--4135--b8ec--6880e9d4d359-outbox (253:19) 1b573d9d--6af4--4135--b8ec--6880e9d4d359-metadata (253:15) 5cf50403--521d--43d8--9318--6229013a1b60-ids (253:11) 5cf50403--521d--43d8--9318--6229013a1b60-metadata (253:9) 5cf50403--521d--43d8--9318--6229013a1b60-master (253:14) 5cf50403--521d--43d8--9318--6229013a1b60-outbox (253:13) 36006048cc786b195adb3de2b6d613cbc (253:8) vg_sf02-lv_home (253:2) 36006048c14e9eb8f668dfc53ea5995ca (253:7) 1b573d9d--6af4--4135--b8ec--6880e9d4d359-ids (253:17) 1b573d9d--6af4--4135--b8ec--6880e9d4d359-leases (253:16) vg_sf02-lv_swap (253:1) 1b573d9d--6af4--4135--b8ec--6880e9d4d359-inbox (253:18) vg_sf02-lv_root (253:0) Version-Release number of selected component (if applicable): rhevm-3.3.0-0.9.master.el6ev.noarch vdsm-4.12.0-rc1.12.git8ee6885.el6.x86_64 How reproducible: Always Steps to Reproduce: 1. Have VM with two disks 2. Create live snapshot 3. Remove VM Actual results: Disks are left on storage domain Expected results: Disks are removed from system Additional info: Worked in 3.2 Failing live snapshot sanity test on 3.3
Are you sure this always reproduced? IMO https://bugzilla.redhat.com/show_bug.cgi?id=978975 might solve this issue on remove failure.
I tried to reproduce the scenario with live snapshot but it didn't reproduced. I think that the issue here is not live snapshot but commit of a previewed snapshot. The reproduced steps which I found from the logs are: 1. Create an images disk in a VM 2. Create a snapshot 3. Create a new image disk 4. preview the snapshot was created 5. Commit the snapshot The last disk which was created has been removed from the DB but it was not removed from the VDSM completely. (The code that should be checked here is _imagesToDelete list which being initialized in the RestoreFromSnapshotCommand and passed as an argument to DestroyImageVdsCommand, probably destroy Image should be called here.) This bug is not a new behaviour since it appears there were always volumes leftovers in the storage, but it was not an issue until we introduced the new disks subtab which we also use the same query probably to find if there are orphaned images when deactivating a storage domain. Regarding the disks which could not be removed, I think that the problem here was that since those disks were already removed from the DB there are no permissions related to them, therefore we can't remove those disks from the storage.
For the record: Maor was right. The correct reproducer is: 1. Create an images disk in a VM 2. Create a snapshot 3. Preview the created snapshot 4. Commit the snapshot
I am removing Regression keyword based on comment 5 of Maor and comment 8 where Kuba agrees that Maor's step to reproduce are correct.
Tested on FCP Data Centers Verified, tested on RHEVM 3.3 - IS18 environment: Host OS: RHEL 6.5 RHEVM: rhevm-3.3.0-0.25.beta1.el6ev.noarch PythonSDK: rhevm-sdk-python-3.3.0.15-1.el6ev.noarch VDSM: vdsm-4.13.0-0.2.beta1.el6ev.x86_64 LIBVIRT: libvirt-0.10.2-27.el6.x86_64 QEMU & KVM: qemu-kvm-rhev-0.12.1.2-2.412.el6.x86_64 SANLOCK: sanlock-2.8-1.el6.x86_64
Closing - RHEV 3.3 Released