Use the following script to mount a persistent storage during startup. This is helpful when you provision the compute instance as well as the persistent disk using some automation tool like Terraform. StackOverFlow question - https://stackoverflow.com/questions/53162620/automate-gcp-persistent-disk-initialization

gcp_compute_mount_persistent_disk.sh

#!/bin/bash
set -uxo pipefail

# DISK_NAME = Name of the disk in terraform
# DEVICE_NAME = When $DISK_NAME is mounted in the compute instance at `/dev/` 

MOUNT_DIR=/mnt/disks/persistent_storage

# Check if entry exists in fstab
grep -q "$MOUNT_DIR" /etc/fstab
if [[ $? -eq 0 ]]; then # Entry exists
	exit
else
	set -e # The grep above returns non-zero for no matches & we don't want to exit then.

	# Find persistent disk's drive value, prefixed by `google-` # https://registry.terraform.io/providers/hashicorp/google/latest/docs/resources/compute_instance#device_name
	DEVICE_NAME="/dev/$(basename $(readlink /dev/disk/by-id/google-${DISK_NAME}))"

	sudo mkfs.ext4 -m 0 -F -E lazy_itable_init=0,lazy_journal_init=0,discard $DEVICE_NAME
	sudo mkdir -p $MOUNT_DIR
	sudo mount -o discard,defaults $DEVICE_NAME $MOUNT_DIR

	# Add fstab entry
	echo UUID=$(sudo blkid -s UUID -o value $DEVICE_NAME) $MOUNT_DIR ext4 discard,defaults,nofail 0 2 | sudo tee -a /etc/fstab
fi

Given that DISK_NAME is under your control, wouldn't you want to use the device path in fstab rather than the UUID? I.e., if the disk is replaced, you'd want the new one with the same name, not the old one by UUID?

@tclift depends on the use case. The thinking here was that if you decide to change the DISK_NAME, the UUID would still be the same so it would mount correctly.

Makes sense. Thanks, I found this useful.

Will your script wipe all data if we destroy the instance and recreate it ?