About 2 weeks after Ubuntu16.04 Spectre security update, the instances in Amazon AWS - started to die. They do not go back to life after reboot, the logs (when disk is attached to other instance) does not show the cause (or I missed it).
I did find out, that changing he instance type from c5 to m4 allows it to boot and work correctly (without any system change).
My guess is, that there is a bug at AWS (c5 is a 'new' processor type). I did spot that my hostname has changed, I've seen some cloud-init reinitialization - but this 'may' be normal (for now, I do run from snapshot).
Going to do some other test - but until now - I'd love to hear did you have similar issue with AWS c5 instances running Ubuntu
Note that the last running c5 instance is in zone eu-west-1c
Interesting - the instance (c5.xlarge) is active for a second during boot:
x:~$ ping prd-pred PING prd-pred (10.0.2.111) 56(84) bytes of data. 64 bytes from 10.0.2.111: icmp_seq=3 ttl=62 time=53.2 ms ^C --- prd-pred ping statistics --- 41 packets transmitted, 1 received, 97% packet loss, time 40312ms rtt min/avg/max/mdev = 53.281/53.281/53.281/0.000 ms
Edit - it does not work with m5.xlarge Was it change from xen to kvm? Or why did it break after some undefined period of time...
Edit my guess is that AWS is doing rolling update of they 5-family instance-supervisors and randomly breaks the systems...