acpid not started on bionic on ec2

[Phil Sweeney]

EC2 instances based on the new nitro hypervisor (e.g. C5/M5 families) need acpid running to cleanly shut down when stopped from the cloud provider's end (e.g. from API call/console).

If I launch the latest Ubuntu 16.04.4 LTS AMI:

$ ps ax | grep acpid
 1141 ? Ss 0:00 /usr/sbin/acpid
 1428 pts/0 S+ 0:00 grep --color=auto acpid

If I launch the latest bionic nightly AMI (ubuntu/images-testing/hvm-ssd/ubuntu-bionic-daily-amd64-server-20180424):

~$ ps ax | grep acpid
 1236 pts/0 S+ 0:00 grep --color=auto acpid

I did some tests, and stopping the instance running 18.04 from the console is a hard (unclean) shut down on an m5 instance. If I instead run this command:

sudo systemctl start acpid

Then stop the instance from the console, I can see Ubuntu shuts down gracefully.

So seems to be a regression in 18.04 Bionic.

[Joshua Powers (powersj)]

Hi,

I launched a recent image (20200908) and can confirm that acpid is running:

$ systemctl status acpid
● acpid.service - ACPI event daemon
   Loaded: loaded (/lib/systemd/system/acpid.service; disabled; vendor preset: enabled)
   Active: active (running) since Fri 2020-10-09 19:55:31 UTC; 54s ago
 Main PID: 840 (acpid)
    Tasks: 1 (limit: 4402)
   CGroup: /system.slice/acpid.service
           └─840 /usr/sbin/acpid

Oct 09 19:55:31 ip-172-31-49-102 acpid[840]: starting up with netlink and the input layer
Oct 09 19:55:31 ip-172-31-49-102 acpid[840]: 2 rules loaded
Oct 09 19:55:31 ip-172-31-49-102 acpid[840]: waiting for events: event logging is off
Oct 09 19:55:31 ip-172-31-49-102 systemd[1]: Started ACPI event daemon.

I am going to close this as fix released. If you disagree or have more information please do feel free to follow-up.

Thanks!