Created attachment 949356 [details] logs Description of problem: The Creation of a new Vm + qcow sparse block image fails during OS installation, it seems that the image "refuses" to grow. 2014-10-22 13:34:29,573 INFO [org.ovirt.engine.core.dal.dbbroker.auditloghandling.AuditLogDirector] (DefaultQuartzScheduler_Worker-85) Correlation ID: 313f94d a, Job ID: 55e34983-d107-4357-a3dc-1c9659b1d165, Call Stack: null, Custom Event ID: -1, Message: VM vm_virtio_thin started on Host ibm-03-ppc 2014-10-22 13:35:49,900 ERROR [org.ovirt.engine.core.vdsbroker.vdsbroker.VdsBrokerObjectsBuilder] (DefaultQuartzScheduler_Worker-39) Error in parsing vm pause status. Setting value to NONE 2014-10-22 13:35:49,903 INFO [org.ovirt.engine.core.vdsbroker.VdsUpdateRunTimeInfo] (DefaultQuartzScheduler_Worker-39) VM vm_virtio_thin 087891ff-eb61-401d-a50a-0665cb43207d moved from Up --> Paused Version-Release number of selected component (if applicable): ppc-ibm How reproducible: 100% Steps to Reproduce: 1.create Vm+block thin provision disk 2.install OS Actual results: The operation fails Expected results: The operation should be successful sparse image should grow Additional info:
I suppose libvirt and qemu logs would be needed too?
it also doesn't look like necessarily related to qcow expansion….it's just 70s after the VM was started… Nir, may it be the missing pause code (as I see none is reported in the vdsm log)?
This looks like a duplicate of bug 1127460, so this depends on 1150015. On rhel7 you cannot use thin provisioning without fixing the udev rules. Since I did not backport the fixes to 3.4, it is expected to fail when installing an os - although not in 70 seconds. Ori, to verify that this is the issue: 1. Start a vm with one disk 2. Find the lv name - should appear in vdsm log in the createVolume log 3. Find the device mapper device backing up the lv /dev/dm-xxx 4. Check the selinux label of the device: ls -Z /dev/dm-xxx Should be something like: system_r:object_r:svirt_image_t:s0:cxxx,cyyy 5. Now start os installation 6. Watch the selinux label during installation 7. When the first disk extend is performed, and the disk size changes from 1G to 2G, the vm will pause 8. The selinux label would be: system_r:object_r:fixed_disk_device_t:s0
Add back needinfo for Ori (thanks bugzilla).
note,this bug is dependent on BZ #1156038, VM's fail to start, hence, couldn't verify the behavior described in #3
Ori, there's a workaround documented in that bug. You cam continue with verification of this bug
Guys, sorry for the long interval :) I've checked several block thin provision disk growth via command line and I confirm that the observed behavior is as noted at comment #3, hence, this is a duplicate of bz #1150015. Michal, However I strongly suggest to verify this bug on powerKVM environment, it is too important to fall between the cracks, your decision.
(In reply to Ori from comment #11) > Guys, sorry for the long interval :) > I've checked several block thin provision disk growth via command line and I > confirm that the observed behavior is as noted at comment #3, hence, this is > a duplicate of bz #1150015. > > Michal, However I strongly suggest to verify this bug on powerKVM > environment, it is too important to fall between the cracks, your decision. Agreed. Moving to QA to verify on PPC.
VM installation went fine using a thin-provision disk located on block storage. Disk size grows on writing to it. verified with latest build_39
RHEV-M 3.5.0 has been released, closing this bug.