There was an event that the OS restarted without shutting down. The shutdown message was not output to syslog.

```
Jul 23 22:55:01 CROND[30576]: (root) CMD (/usr/lib64/sa/sa1 1 1)
Jul 23 22:56:09 kernel: imklog 5.8.10, log source = /proc/kmsg started.
Jul 23 22:56:09 rsyslogd: [origin software="rsyslogd" swVersion="5.8.10" x-pid="1704" x-info="http://www.rsyslog.com"] start
Jul 23 22:56:09 kernel: Initializing cgroup subsys cpuset
Jul 23 22:56:09 kernel: Initializing cgroup subsys cpu
```

I thought that the physical server was powered off, so I contacted the cloud vendor.

However, they answered that the physical server was not powered off.

This server is in a cluster configuration using Pacemaker. I thought that Pacemaker fencing worked.

But, Pacemaker's fencing seems to be disabled

```
stonith-enabled="false"
```

What can be considered as the cause of the OS reboot?

kernel 2.6.32-431.el6
Distribution CentOS 6.6
Heartbeat 3.0.5-1.1.el6
Pacemaker 1.1.18-3.el6

What does show?

thank you for your comment.

The command execution result is as follows↓
=================================================================================
$ last | grep boot

reboot system boot 2.6.32-431.el6.x Wed Jul 23 22:55 - 06:02 (1+08:06)
=================================================================================

so the system was rebooted at Wed Jul 23 22:55 - 06:02 (1+08:06)

I think you now need to look in the messages log to see what, if anything, happened just before then, but I'm not sure you can tell what happened. When everything goes away suddenly logging stops (doh).

Perhaps you're not asking the right question of your vendor. There are many reasons besides a power off that will cause a VM to reboot. You should open a ticket with them asking "Why did my VM reboot on Wed Jul 23 22:55 - 06:02 (1+08:06)"? At a minimum, you'll make them aware it happened, and they may be able to provide an RFO.

Was the server off-line for a significant time? Were your services affected?
Don't tell us...put the answers to those questions in your support ticket.

Vendors hypervisor ran out of RAM and needed to start killing processes?

I saw pacemaker freeze kernel if it cannot fence. I did not find any explanation of that behavior.
Once kernel frozen, any running watchdog can reset VM. To prove that, you can ask your provider for this kind event occur.