Hi Markus, Can you try to use dd on a random file in that storage domain, What is the performance it reflects? Just want to make sure what is the origin of the problem, if it is related to the storage domain or the qemu process. Regards, Maor On Thu, Nov 3, 2016 at 8:09 PM, Markus Stockhausen <stockhausen@collogia.de> wrote:
Hi there,
in the past we already observed that live merge operations may take some time. Currently another of those operations is running on one of our VMs on a CentOS 7/qemu 2.3 node. Storage is (async - yeah I know about potential data loss) NFS over 10GBit network.
The operation is running for 50 minutes now with an average throughput of 5MByte/sec. The disk (defined at 400GB) and its snapshots does not show much thin provisioning:
ls -al -rw-rw----. 1 vdsm kvm 429496729600 7. Apr 2016 a204865e-ef57-41e0-a2ac-465a6c9c3d60 -rw-rw----. 1 vdsm kvm 203334418432 3. Nov 2016 aeda833f-1f4b-4b6f-b553-578f1f3f06ee -rw-rw----. 1 vdsm kvm 192091389952 3. Nov 2016 c8acdbc7-af24-4c5c-94c5-ae7262d98f5c
du -m a204865e-ef57-41e0-a2ac-465a6c9c3d60 383070 a204865e-ef57-41e0-a2ac-465a6c9c3d60 du -m aeda833f-1f4b-4b6f-b553-578f1f3f06ee 193953 aeda833f-1f4b-4b6f-b553-578f1f3f06ee du -m c8acdbc7-af24-4c5c-94c5-ae7262d98f5c 183222 c8acdbc7-af24-4c5c-94c5-ae7262d98f5c
Usually at some point the process is gaining speed and we see >100MByte/sec speed. Can anyone explain what might be going on.
Best regards.
Markus Stockhausen
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users