mihajlov | xnox, as a followup to the discussion you had with macjl, guest migration between different KVM versions needs more consideration than the security driver used | 11:15 |
---|---|---|
mihajlov | e.g. the target hypervisor must understand and support the machine type as defined on the source hypervisor | 11:17 |
mihajlov | a guest defined on KVM for IBM z will by default have a machine type of s390-ccw-kvmibm-1.1.1 | 11:19 |
mihajlov | which won't be accepted by an upstream QEMU | 11:20 |
mihajlov | one could define the guest with machine type of s390-ccw-virtio-2.4 on the KVM for IBM z hypervisor which would allow the migration to an upstream hypervisor (like Ubuntu) | 11:22 |
xnox | mihajlov, right. | 11:23 |
xnox | mihajlov, forgot that part. When I saw this different machine type, i was slightly surprised and confused why a new one got defined. | 11:24 |
xnox | mihajlov, I can certainly take that machine type as a patch to Debian & Ubuntu qemu, if we do in fact support / are compatible two way between 2.5 and e.g. s390-ccw-kvmibm-1.1.1 | 11:24 |
mihajlov | regarding the selinux <-> apparmor conversion I have doubts whether an automatic conversion can be vouched to be safe | 11:25 |
xnox | true | 11:25 |
mihajlov | as a potential way out, it is possible to send a modified domain XML over to the target machine using the --xml option on the migrate command | 11:26 |
xnox | mihajlov, i wonder if we can, and/or should enable selinux in qemu on ubuntu. we have selinux anabled in a bunch of things | 11:26 |
mihajlov | where you could omit the security driver | 11:26 |
xnox | (and e.g. smack too) | 11:26 |
xnox | mihajlov, would z/kvm accept --to-ubuntu flag? =) | 11:26 |
* xnox is biased and wants everything on ubuntu ;-) | 11:27 | |
mihajlov | xnox, wrt to selinux on Ubuntu I am not a security expert but I thought you'll have to chose one method for your system? | 11:28 |
xnox | apparmor is default and integrated throughout. | 11:29 |
xnox | however other systems are available too, for those that want to use them. | 11:29 |
xnox | e.g. we had selinux enabled in upstart as pid 1, because there are selinux usecases that people resonably use. | 11:29 |
mihajlov | xnox, regarding the flag: to have one would be less of a problem (of course it would be one upstream) then the semantic associated with it | 11:29 |
xnox | --insecure or some such =) | 11:30 |
mihajlov | there's no way to "downgrade" a running virtual machine without impacting the guest | 11:30 |
xnox | i dunno if libvirtd can do multiple security models simultaniously. | 11:30 |
xnox | ouch. | 11:30 |
mihajlov | would be a matter of testing I think | 11:35 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!