virsh_migrate_stress: perform migration with macvtap hotplug/coldplug #2064
Conversation
|
depends on: avocado-framework/avocado-vt#1985 |
balamuruhans
force-pushed
the
macvtap_migration
branch
from
35a8562 to
fa55625
Compare
balamuruhans
force-pushed
the
macvtap_migration
branch
from
fa55625 to
03c9210
Compare
|
depends on: avocado-framework/avocado-vt#2042 |
|
@dzhengfy please help to review. |
| pf_vf_info = source_assignable.get_pf_vf_info() | ||
| # map vf from each pf to every VM | ||
| for vm_index in range(len(vms)): | ||
| iface_list = [] |
There was a problem hiding this comment.
check if there is a possibility of abstracting to config like if applicable
- one vfs from all pfs to all vms
- all vfs of one pf to each vm( helpful incase of single vm)
- all vfs of all pfs to one vm
There was a problem hiding this comment.
I will add these as new scenarios in my next PR.
| vstress[vm.name].unload_stress() | ||
| except KeyError: | ||
| continue | ||
| # bring down the VMs |
There was a problem hiding this comment.
I guess we can skip and leave it postprocess based on the config setting?, what say
There was a problem hiding this comment.
So, these tests would run stress tools inside VM, bringing down the VMs here will actually cleans up all of it, because sometimes after test completes it takes more time for VM to respond in login, kill the stress process and clean up due to the stress load running inside VM. It is better not to waste test time instead bring down the VM will helps us save it.
balamuruhans
force-pushed
the
macvtap_migration
branch
from
03c9210 to
73ccf38
Compare
|
@sathnaga I have made the changes as suggested. |
This patch adds tests that could perform migration with macvtap interface attached to VMs. pre-requisite is to have physical cards attached in the same slots in source and target host so that the Virtual Functions(VFs) created out of Physical Functions(PFs) will be identical for VMs migrate. This patch removes couple of trivial scenarios that can be tested by overriding the `smp` and `mem` params from config instead. Signed-off-by: Balamuruhan S <[email protected]>
balamuruhans
force-pushed
the
macvtap_migration
branch
from
73ccf38 to
61952bf
Compare
|
After confirmation with the IBM PPC FVT lead that the author does not work on this any more and the pr is not in use by them. So I close it. |
This patch adds tests that could perform migration with macvtap
interface attached to VMs. pre-requisite is to have physical cards
attached in the same slots in source and target host so that the
Virtual Functions(VFs) created out of Physical Functions(PFs) will
be identical for VMs migrate.
This patch removes couple of trivial scenarios that can be tested
by overriding the
smpandmemparams from config instead.Signed-off-by: Balamuruhan S [email protected]