Feature #1484
closed
libvirt: map rbd via kernel driver
Added by Sage Weil over 12 years ago.
Updated over 7 years ago.
Description
this should share a common image description schema as the qemu driver. we can treat qemu+librbd as a (preferred) optimization, as long as we have an option to force use of the kernel driver or librbd (and have the latter fail on non-qemu hypervisors). Something along the lines of type = {krbd, librbd, auto}.
- Target version deleted (
v0.37)
- Translation missing: en.field_position set to 1
- Translation missing: en.field_position deleted (
13)
- Translation missing: en.field_position set to 12
- Translation missing: en.field_position deleted (
13)
- Translation missing: en.field_position set to 12
- Translation missing: en.field_position deleted (
46)
- Translation missing: en.field_position set to 13
In the current design of libvirt I don't see how you could achieve this.
With my storage pool work I found out that you can only start and stop an entire storage pool inside libvirt.
This would be you would (un)map all images inside a specific pool, you can't start (map) or stop (unmap) just one image.
If you are talking about mapping an image to a virtual machine, I'm not aware of any hooks you can execute, but why would you? You can pass down /dev/rbd0 as a device to a guest, you just need some way to get it mapped, that's the hardest part.
- Project changed from Ceph to rbd
- Category deleted (
libvirt)
- Translation missing: en.field_position deleted (
319)
- Translation missing: en.field_position set to 1326
- Status changed from New to Rejected
Not a high-priority objective for RBD
Also available in: Atom
PDF