Script to create swap partition fails when running automatically

Question

I am creating a cluster of machines in AWS (Amazon Linux 2) using terraform by utilizing the user_data argument under the aws_instance resource. Part of the script is the creation of a swap partition. The commands in my script work perfectly if I execute them manually in the instance.

I have tried the following in my script. It creates the partition successfully, but it does not seem to finish up creating the swap as confirmed in more /proc/swaps. It successfully executes the lines of code below everything I have showed that I omitted from my post. So it must be failing at the partprobe, mkswap /dev/nvme1n1p2, or swapon /dev/nvme1n1p2. It runs the echo "/dev/nvme1n1p2 swap swap defaults 0 0" >> /etc/fstab. I'm not sure how to tell where it is not executing.

# Create SWAP partition
fdisk /dev/nvme1n1 <<EOF
n
p
2

+48G
w
EOF
partprobe
mkswap /dev/nvme1n1p2
swapon /dev/nvme1n1p2
echo "/dev/nvme1n1p2 swap swap defaults 0 0" >> /etc/fstab

The results intended are to create swap partition as confirmed by running more /proc/swaps. The partition is created, but not the actual swap.

UPDATE: Here is the contents of the log file:

mkswap: cannot open /dev/nvme1n1p2: No such file or directory
swapon: cannot open /dev/nvme1n1p2: No such file or directory

However that device is listed when running lsblk and the command works if I run it manually.

Answer 1

Solution from the comments by @thatotherguy:

I'm guessing it's a race condition with partprobe. If you run manually, there will be several seconds between each command so udev will have plenty of time to create device node asynchronously. When you run them in a script, it doesn't. You could try adding sleep 5 after partprobe.

This solved the OP's error, and I can also report success using this fix.