Running dracut on AWS EBS Centos 7 prevents booting

Question

I have a minimal Centos 7 build. When I run dracut --force, it builds a new initramfs that is 20mb smaller. The only difference in the modules section is an absence of qemu.

localhost login: [ 3322.193617] Restarting system.
[    0.000000] Cannot get hvm parameter CONSOLE_EVTCHN (18): -22!
[    0.817440] Cannot get hvm parameter CONSOLE_EVTCHN (18): -22!
[[32m  OK  [0m] Started Show Plymouth Boot Screen.
[[32m  OK  [0m] Reached target Paths.
[[32m  OK  [0m] Reached target Basic System.
[  123.897756] dracut-initqueue[273]: Warning: dracut-initqueue timeout - starting timeout scripts
[  124.410440] dracut-initqueue[273]: Warning: dracut-initqueue timeout - starting timeout scripts

Answer 1

The boot process for an HVM EC2 instance appears to still require xen-blkfront. However, these are not detected when dracut rebuilds, nor do they show up with modprobe on the running system. I added them using the example in this post and it successfully rebooted.

Answer 2

If you want to have more generic initramfs to support various types of AWS instance, such as modern M5/C5/t3a, enhanced networking adapter (ENA), EBS optimized disk, nvme, generic xen-blkfront ... etc... You should disable dracut option host-only. (In CentOS default is on)

In short:

dracut -f -v -N

OR

yum install dracut-config-generic
dracult -f -v

Double check: (you should see drivers for ena, xen, nvme, xen-blkfront, etc.)

lsinitrd | grep "ena\|xen\|nvme"

And remember to make sure your /etc/fstab using UUID to map partition instead of device name such as /dev/sdx