mjhjmtu

Context

My product provisions spot instances on AWS. The instances should appear identical after a user terminates, and then recreates that instance (we need to remount the volume). AWS suggests temporarily shutting down instances, and then swapping the volume via their API; unfortunately, they don't support this for all combinations of spot pricing & instance types.

NB: Depending on instance type, when you attach /dev/sdf it may show up as /dev/nvme1n1p1 instead.

What I've Tried

/sbin/init doesn't appear to do anything after running this script. I tried exec switch_root /next_root /sbin/init at some point to tidy things but that didn't work either.

sudo mkdir -p /next_root
sudo rm /sbin/init

# if the output of lsblk contains "nvme1n1p1"
# then DEVICE="nvme1n1p1"
# else DEVICE="sdf"
DEVICE="sdf" && [[ $(lsblk) = *"nvme1n1p1"* ]] && DEVICE="nvme1n1p1"
DEVICE="/dev/$DEVICE"

sudo e2label $DEVICE next_root
sudo tune2fs $DEVICE -U `uuidgen`

sudo tee /sbin/init > /dev/null <<EOF
#!/bin/sh

mount $DEVICE /next_root

cd /next_root
mkdir -p ./old_root
pivot_root . ./old_root

mount --move /sys ./sys
mount --move /proc ./proc
mount --move /dev ./dev
mount --move /run ./run

exec chroot . /sbin/init
EOF

sudo chmod +x /sbin/init
sudo telinit u

I also tried pivot_root as su but stderr reported pivot_root: failed to change root from '.' to './old_root': Invalid argument

# ...

sudo su - <<EOF
mount $DEVICE /next_root

cd /next_root
mkdir -p ./old_root
pivot_root . ./old_root

# ...

exec chroot . /sbin/init
EOF

I've been stumped on this one for a few days and would really appreciate your help!

Thank you for reading.