[QEMU] Fixing an unintentional variable shadowing

[simar7]

This PR fixes the shadowing of the monitorPath variable
within the QEMU driver for Nomad. With this variable being shadowed
within the scope, all graceful shutdowns are effectively disabled
as monitorPath will not be set.

Also adds a helpful debug statement for the operator to verify.

Signed-off-by: Simarpreet Singh simar@linux.com

[schmichael]

Nice catch!

[schmichael]

It's nice to include something to indicate what task log lines refer to. Perhaps something like:

[DEBUG] driver.qemu: graceful shutdown disabled, sending interrupt signal to pid %d", h.userPid

[simar7]

@schmichael do we know if the tests are broken on master? The changes I've made don't seem to be related to the changes in the PR.

[dadgar]

@simar7 Can we improve the test to check that the vm shutdown post signaling? The test should fail before this change and pass after.

[cheeseprocedure]

Chiming in, as I investigated validation of VM shutdown as part of PR# 3411: the minimal Linux image in linux-0.2.img doesn't appear to respond to an ACPI shutdown signal passed via the QEMU monitor, which makes a full test more difficult. (This can surely be fixed, but I haven't dug any deeper.)

[schmichael]

@cheeseprocedure @simar7 I tried adding an Alpine Linux image in #3819 to test graceful shutdowns and have had no luck. Not sure what I'm doing wrong.

(I pushed it to my own fork to keep the 40MB image out of the main Nomad repo if this effort doesn't work out.)

[cheeseprocedure]

@schmichael it looks like TestQemuDriver_GracefulShutdown does pass using the Alpine image if you add a significant time.Sleep after confirming the presence of the monitor socket. I'm guessing this is because the guest has to reach a state where it's ready to actually receive the ACPI-poweroff signal.

I'm not aware of a sensible way to verify the state of the guest in a black-box fashion before triggering a shutdown. Perhaps we could call sendQemuShutdown on a ticker until the test times out or the process exits?

[schmichael]

@cheeseprocedure Thanks for the tip! The Alpine image has sshd installed, so I just poll until sshd starts and then know it's safe to issue the shutdown command. Pushed updates to #3819.

Looks like everything is working!

[github-actions]

I'm going to lock this pull request because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active contributions.
If you have found a problem that seems related to this change, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.