13:32:18 <al_s> Hi.  So, I have a few omnios instances in AWS running on nitro instance types.   Today they all, at some point or other, rebooted with no prompting from me.  I cannot see any RebootInstances API calls in cloudtrail, so I am reasonably sure no one initiated a reboot from the AWS console or with awscli.  They all show log messages like "Jul 19
13:32:19 <al_s> 12:26:47 ip-xx-xx-xx-xx poweroff: [ID 330035 auth.crit] initiated by LOGIN on /dev/console"   All machines have the console on the serial port.   I remember from many moons ago sparc solaris boxes rebooting when the machine at the other end of the serial cable was rebooted.  Is this still a thing on omnios on x86?   If so, does it seem
13:32:19 <al_s> possible that a similar reboot could be triggered in AWS?
14:46:17 <ptribble> That's the message from someone pressing the (virtual in this case) power button.
14:47:31 <ptribble> Which sounds like something associated with the AWS host underneath the instance.
14:55:48 <al_s> Well, yes.  But that should be in response to a RebootInstances api call, that is what I see when I reboot from the console.  But there were none.
14:56:16 <al_s> reboot from the AWS console
14:57:43 <ptribble> No, if the AWS host underneath reboots, there is no API call at the instance level.
15:03:22 <al_s> Ok.  But all of them in one day?  It was a very odd event.  A number of them rebooted more than once.  No notifications from AWS.   I ran "svccfg -s svc:/system/keymap:default setprop keymap/keyboard_abort=disabled && kbd -i" on them all just in case the serial break was to blame.
15:04:41 <ptribble> That coincidence is more like an underlying infrastructure event - certainly when I was running lots of OmniOS on AWS the unexplained reboots tended to come in clusters.
15:06:21 <al_s> One interesting thing to note is that none of the t2 instances rebooted during the event.
15:21:00 <al_s> Another interesting thing to note is none of the linux t3 instances rebooted.
15:21:17 <al_s> And there are many more of those.