Creating Thin-provisioned disks on a gluster volume
Hi - We want to migrate our aging oVirt cluster to a new one and decided to go for the hyperconverged setup with Gluster and hosted engine. Got 3 new machines, set everything up, did a test VM migration and were almost ready to do final migration of our VMs when I notice one really weird thing. When migrating VMs from NFS storage onto Gluster they stay thin provisioned. But if I try creating new disk image from the oVirt GUI - I get preallocated volume. No matter if I select "thin provision" or "Preallocated" I always get preallocated volume created. The engine.log says - 2019-05-07 17:14:20,562-04 INFO [org.ovirt.engine.core.vdsbroker.irsbroker.CreateImageVDSCommand] (default task-20) [bbd94792-2f86-44eb-9334-807dd44d0742] START, CreateImageVDSCommand( CreateImageVDSCommandParameters:{storagePoolId='43550c2c-7100-11e9-9b45-00163e4babb7', ignoreFailoverLimit='false', storageDomainId='1de2f7ac-ee52-43fb-89ac-f86b62f73b9a', imageGroupId='86c5c5e9-6533-4caf-a5a5-401556a5059a', imageSizeInBytes='21474836480', volumeFormat='RAW', newImageId='470b8035-9380-408d-a644-96e8d74c4e8e',* imageType='Preallocated', *newImageDescription='{"DiskAlias":"test","DiskDescription":""}', imageInitialSizeInBytes='0'}), log id: 228ddfd5 Is there any tweak I didn't notice to make things actually what it suppose to be doing? oVirt 4.3.3.5-1.el7 glusterfs 5.5-1.el7 CentOS Linux release 7.6.1810 Thank you! -- Dmitry Filonov Linux Administrator SBGrid Core | Harvard Medical School 250 Longwood Ave, SGM-114 Boston, MA 02115
Il Mar 7 Mag 2019, 23:50 Dmitry Filonov <filonov@hkl.hms.harvard.edu> ha scritto: [Snip]
But if I try creating new disk image from the oVirt GUI - I get preallocated volume. No matter if I select "thin provision" or "Preallocated" I always get preallocated volume created.
The engine.log says - 2019-05-07 17:14:20,562-04 INFO [org.ovirt.engine.core.vdsbroker.irsbroker.CreateImageVDSCommand] (default task-20) [bbd94792-2f86-44eb-9334-807dd44d0742] START, CreateImageVDSCommand(
[Snip]
*imageType='Preallocated', *
[Snip]
Is there any tweak I didn't notice to make things actually what it suppose to be doing?
oVirt 4.3.3.5-1.el7 glusterfs 5.5-1.el7 CentOS Linux release 7.6.1810
Thank you!
It seems this is the third separate thread regarding the same problem since 25/04 but I have not seen any answer (positive or negative) from one of the oVirt developers / maintainers yet on the list... (perhaps right now probably involved in Red Hat Summit.. ) Anyway I see that Strahil already opened a bug here: https://bugzilla.redhat.com/show_bug.cgi?id=1704782 And there is useful information inside it, such as that the problem is only on gluster and only if disk is created from webadmin GUI, while from rest API is ok. Also from 4.2 to 4.3 disk format (only on gluster) changed default from thin provisioned to preallocated. Hih, Gianluca
Thank you, Gianluca! I was searching list archives but couldn't find anything recent. All threads were from years ago. Glad am not the only one. Yes, the problem is only from GUI and gluster only. Thanks again. Fil On Tue, May 7, 2019, 6:30 PM Gianluca Cecchi <gianluca.cecchi@gmail.com> wrote:
Il Mar 7 Mag 2019, 23:50 Dmitry Filonov <filonov@hkl.hms.harvard.edu> ha scritto: [Snip]
But if I try creating new disk image from the oVirt GUI - I get preallocated volume. No matter if I select "thin provision" or "Preallocated" I always get preallocated volume created.
The engine.log says - 2019-05-07 17:14:20,562-04 INFO [org.ovirt.engine.core.vdsbroker.irsbroker.CreateImageVDSCommand] (default task-20) [bbd94792-2f86-44eb-9334-807dd44d0742] START, CreateImageVDSCommand(
[Snip]
*imageType='Preallocated', *
[Snip]
Is there any tweak I didn't notice to make things actually what it suppose to be doing?
oVirt 4.3.3.5-1.el7 glusterfs 5.5-1.el7 CentOS Linux release 7.6.1810
Thank you!
It seems this is the third separate thread regarding the same problem since 25/04 but I have not seen any answer (positive or negative) from one of the oVirt developers / maintainers yet on the list... (perhaps right now probably involved in Red Hat Summit.. ) Anyway I see that Strahil already opened a bug here: https://bugzilla.redhat.com/show_bug.cgi?id=1704782 And there is useful information inside it, such as that the problem is only on gluster and only if disk is created from webadmin GUI, while from rest API is ok. Also from 4.2 to 4.3 disk format (only on gluster) changed default from thin provisioned to preallocated. Hih, Gianluca
participants (2)
-
Dmitry Filonov -
Gianluca Cecchi