qemu_guest_agent: Add new api support 'guest-get-load' #4309
Conversation
6-dehan
force-pushed
the
KVMAUTOMA-3534-cpuload
branch
7 times, most recently
from
1a6487d to
c9269f4
Compare
|
for linux guest: |
6-dehan
force-pushed
the
KVMAUTOMA-3534-cpuload
branch
3 times, most recently
from
5ca1445 to
fd0cfe1
Compare
|
For Window VM: |
|
@leidwang Could you please help to review this when you're free? |
| # Start CPU stress test | ||
| error_context.context("Start CPU stress test", LOG_JOB.info) | ||
| session.cmd(params["cmd_run_stress"]) | ||
| time.sleep(10) |
There was a problem hiding this comment.
Hi @6-dehan
I saw you added some time.sleep to wait for the process to start or terminate, I'm worried it will introduce some failures in the future, is it possible to check the process status directly instead of time.sleep?
There was a problem hiding this comment.
Hi @leidwang
I tried to detect the running status, but it is not enough to just have the running status. In the end, we still need to run the stress command for a while to let the stress value rise to meet our expectations. And it is just a simple stress test command. time.sleep is enough to deal with it. There is no need to make it too complicated.
qemu/tests/qemu_guest_agent.py
Outdated
| except (KeyError, ValueError) as e: | ||
| test.error(f"Failed to get QGA load stats: {e}") | ||
|
|
||
| def _verify_load_values(qga_vals, guest_vals, check_type="match"): |
There was a problem hiding this comment.
| def _verify_load_values(qga_vals, guest_vals, check_type="match"): | |
| def _verify_load_values(qga_vals, guest_vals, check_type="match", prev_values=None): |
Add a new guest agent command 'guest-get-load' to get cpu load average info of Guest. Signed-off-by: Dehan Meng <[email protected]>
6-dehan
force-pushed
the
KVMAUTOMA-3534-cpuload
branch
from
7a7c8c3 to
f8a677b
Compare
|
Linux could be passed normally, Windows still has to adjust a little. |
Add a new guest agent command 'guest-get-load' to
get cpu load average info of Guest.
ID: 3534
Signed-off-by: Dehan Meng [email protected]