Server Overlay and Serverless
Can application layer isolated from OS layer?
Overlay
Docker uses overlay system, but it requires docker file to rebulid docker image. My colleague was asking me about the image patching and I told him that upper layer can overwrite the lower layer, there is no way to prevent this.
Layering
If the layering can be done in managed way, then it could be a better option to deploy applications. So the ideal is to separate platform to more layers as below
- OS layer - Kernal and all OS packages are sitting in this layer
- OS configuration layer - This layer consists of operating system configurations, such as network configuration, application filesystem configuration, etc.
- OS to APP Patch layer - Some specific OS requirements for specific application.
- Middleware layer - This includes middleware packages
- Data layer - This includes all data required to run application
- Application layer - This is the actual application
Docker or Containerd
The Docker or Containerd packages can be in OS layer, but the actual configuration should be Middleware layer.
CoreOS
The CoreOS implementation is a good example for this layering.
Sample implementation
solve raspbian SD card corruption issues with read-only mounted root partition
Steps
- Copy the script in next section to
/sbin/overlayRoot.sh and make it executable
sudo chmod +x /sbin/overlayRoot.sh
sudo dphys-swapfile swapoff
sudo dphys-swapfile uninstall
sudo update-rc.d dphys-swapfile remove
- Add following line to the end of
cmdline.txt file in the boot partition:
init=/sbin/overlayRoot.sh
Script
The copy of script is listed below
#!/bin/sh
# Read-only Root-FS for Raspian using overlayfs
# Version 1.1
#
# Version History:
# 1.0: initial release
# 1.1: adopted new fstab style with PARTUUID. the script will now look for a /dev/xyz definiton first
# (old raspbian), if that is not found, it will look for a partition with LABEL=rootfs, if that
# is not found it look for a PARTUUID string in fstab for / and convert that to a device name
# using the blkid command.
#
# Created 2017 by Pascal Suter @ DALCO AG, Switzerland to work on Raspian as custom init script
# (raspbian does not use an initramfs on boot)
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see
# <http://www.gnu.org/licenses/>.
#
#
# Tested with Raspbian mini, 2018-10-09
#
# This script will mount the root filesystem read-only and overlay it with a temporary tempfs
# which is read-write mounted. This is done using the overlayFS which is part of the linux kernel
# since version 3.18.
# when this script is in use, all changes made to anywhere in the root filesystem mount will be lost
# upon reboot of the system. The SD card will only be accessed as read-only drive, which significantly
# helps to prolong its life and prevent filesystem coruption in environments where the system is usually
# not shut down properly
#
# Install:
# copy this script to /sbin/overlayRoot.sh, make it executable and add "init=/sbin/overlayRoot.sh" to the
# cmdline.txt file in the raspbian image's boot partition.
# I strongly recommend to disable swapping before using this. it will work with swap but that just does
# not make sens as the swap file will be stored in the tempfs which again resides in the ram.
# run these commands on the booted raspberry pi BEFORE you set the init=/sbin/overlayRoot.sh boot option:
# sudo dphys-swapfile swapoff
# sudo dphys-swapfile uninstall
# sudo update-rc.d dphys-swapfile remove
#
# To install software, run upgrades and do other changes to the raspberry setup, simply remove the init=
# entry from the cmdline.txt file and reboot, make the changes, add the init= entry and reboot once more.
fail(){
echo -e "$1"
/bin/bash
}
# load module
modprobe overlay
if [ $? -ne 0 ]; then
fail "ERROR: missing overlay kernel module"
fi
# mount /proc
mount -t proc proc /proc
if [ $? -ne 0 ]; then
fail "ERROR: could not mount proc"
fi
# create a writable fs to then create our mountpoints
mount -t tmpfs inittemp /mnt
if [ $? -ne 0 ]; then
fail "ERROR: could not create a temporary filesystem to mount the base filesystems for overlayfs"
fi
mkdir /mnt/lower
mkdir /mnt/rw
mount -t tmpfs root-rw /mnt/rw
if [ $? -ne 0 ]; then
fail "ERROR: could not create tempfs for upper filesystem"
fi
mkdir /mnt/rw/upper
mkdir /mnt/rw/work
mkdir /mnt/newroot
# mount root filesystem readonly
rootDev=`awk '$2 == "/" {print $1}' /etc/fstab`
rootMountOpt=`awk '$2 == "/" {print $4}' /etc/fstab`
rootFsType=`awk '$2 == "/" {print $3}' /etc/fstab`
echo "check if we can locate the root device based on fstab"
blkid $rootDev
if [ $? -gt 0 ]; then
echo "no success, try if a filesystem with label 'rootfs' is avaialble"
rootDevFstab=$rootDev
rootDev=`blkid -L "rootfs"`
if [ $? -gt 0 ]; then
echo "no luck either, try to further parse fstab's root device definition"
echo "try if fstab contains a PARTUUID definition"
echo "$rootDevFstab" | grep 'PARTUUID=\(.*\)-\([0-9]\{2\}\)'
if [ $? -gt 0 ]; then
fail "could not find a root filesystem device in fstab. Make sure that fstab contains a device definition or a PARTUUID entry for / or that the root filesystem has a label 'rootfs' assigned to it"
fi
device=""
partition=""
eval `echo "$rootDevFstab" | sed -e 's/PARTUUID=\(.*\)-\([0-9]\{2\}\)/device=\1;partition=\2/'`
rootDev=`blkid -t "PTUUID=$device" | awk -F : '{print $1}'`p$(($partition))
blkid $rootDev
if [ $? -gt 0 ]; then
fail "The PARTUUID entry in fstab could not be converted into a valid device name. Make sure that fstab contains a device definition or a PARTUUID entry for / or that the root filesystem has a label 'rootfs' assigned to it"
fi
fi
fi
mount -t ${rootFsType} -o ${rootMountOpt},ro ${rootDev} /mnt/lower
if [ $? -ne 0 ]; then
fail "ERROR: could not ro-mount original root partition"
fi
mount -t overlay -o lowerdir=/mnt/lower,upperdir=/mnt/rw/upper,workdir=/mnt/rw/work overlayfs-root /mnt/newroot
if [ $? -ne 0 ]; then
fail "ERROR: could not mount overlayFS"
fi
# create mountpoints inside the new root filesystem-overlay
mkdir /mnt/newroot/ro
mkdir /mnt/newroot/rw
# remove root mount from fstab (this is already a non-permanent modification)
grep -v "$rootDev" /mnt/lower/etc/fstab > /mnt/newroot/etc/fstab
echo "#the original root mount has been removed by overlayRoot.sh" >> /mnt/newroot/etc/fstab
echo "#this is only a temporary modification, the original fstab" >> /mnt/newroot/etc/fstab
echo "#stored on the disk can be found in /ro/etc/fstab" >> /mnt/newroot/etc/fstab
# change to the new overlay root
cd /mnt/newroot
pivot_root . mnt
exec chroot . sh -c "$(cat <<END
# move ro and rw mounts to the new root
mount --move /mnt/mnt/lower/ /ro
if [ $? -ne 0 ]; then
echo "ERROR: could not move ro-root into newroot"
/bin/bash
fi
mount --move /mnt/mnt/rw /rw
if [ $? -ne 0 ]; then
echo "ERROR: could not move tempfs rw mount into newroot"
/bin/bash
fi
# unmount unneeded mounts so we can unmout the old readonly root
umount /mnt/mnt
umount /mnt/proc
umount /mnt/dev
umount /mnt
# continue with regular init
exec /sbin/init
END
)"