[SOLVED] qm importdisk kills my SAN

Aug 2, 2018
5
0
1
47
Hi!

I am migrating from XenServer to PVE. When running qm importdisk with raw disk images on my node local disk it kills my SAN connection. Which is 10GBit/s fibre multipath scsi to Fujitsu Eternus SAN. Vms shows sda: abort (with virtio_scsi) and some SATA emulated vms show i/o error and things goes bad.

EDIT: To clarify raw image sits on my nodes local disk and it is supposed to go to my storage which is located on SAN

Tried to run that import with ionice -c3 but did not much help..

Any ideas why qm importdisk is taking SO MUCH bandwith and kills all my vms?
 
you cannot use lvm-thin in a shared manner, this will not work and destroy your data

if you want to use a shared scsi use normal LVM

That's why I asked :-D

There are a lot of settings you can tweak to optimize iSCSI, offloading, jumbo frames, multipathing, etc. ... the best one I always use is not to use iSCSI at all. I've never had I/O timeouts with FC-based environments.
 
Oh nice to hear now that I cannot use it in shared manner.. well i'll create new LVM pool for that then. I managed to figure this out by wrapping qemu-img into ionice -c2 shell script. Now it does not kill my SAN connection. Thanks for this information and I will use normal LVM when creating bigger data pools for the upcoming cluster.

Have to say that when using now PVE at work and at home it is far best virtualization environment I've ever used. Compared to XenServer its like day and night. Thanks for this support and this awesome software! Marking as solved.