This guide is a "fork" of the eponymous Articles about CentOS 5.9, and takes into account the features of the new OS. There is currently no official Centos8 image from centos.org in AWS Marketplace.
As you know, in the Amazon cloud, virtual instances are launched based on images (the so-called AMI). Amazon provides a large number of them, you can also use public images prepared by third parties, for which the cloud provider, of course, does not bear any responsibility. But sometimes you need an image of a clean system with the right parameters, which is not in the list of images.
Then the only way out is to make your own AMI.
The official documentation describes way creating an "instance store-backed AMI".
The disadvantage of this approach is that the finished image will also need to be converted into an βEBS-backed AMIβ. Also worth noting is the Cockpit Image Builder. It will allow you to create custom images, in CLI or WEB GUI mode, but when you already have Centos 8.
How to create your own EBS-backed AMI in the Amazon cloud without intermediate steps will be discussed in this article.
Action plan
- Prepare environment
- Install a clean system, make the necessary settings
- Take a snapshot of the disk
- Register AMI
Preparing the Environment
For our purposes, any official Centos 7 instance any shape, even t2.micro. You can start it via CLI:
aws ec2 run-instances
--image-id ami-4bf3d731
--region us-east-1
--key-name alpha
--instance-type t2.micro
--subnet-id subnet-240a8618
--associate-public-ip-address
--block-device-mappings DeviceName=/dev/sda1,Ebs={VolumeSize=8}
--block-device-mappings DeviceName=/dev/sdb,Ebs={VolumeSize=4}
The command will raise an instance in the VPC to which the specified subnet-id belongs. The subnet is supposed to be public and SG 'default' allows everything.
Now log in to the instance via ssh, update the system, install dnf
and reboot:
sudo yum update -y && sudo yum install -y dnf && sudo reboot
All further operations will be performed from root
.
Installing clean Centos 8.1
File system layout and partition mounting
DEVICE=/dev/xvdb
ROOTFS=/rootfs
parted -s ${DEVICE} mktable gpt
parted -s ${DEVICE} mkpart primary ext2 1 2
parted -s ${DEVICE} set 1 bios_grub on
parted -s ${DEVICE} mkpart primary xfs 2 100%
mkfs.xfs -L root ${DEVICE}2
mkdir -p $ROOTFS
mount ${DEVICE}2 $ROOTFS
mkdir $ROOTFS/{proc,sys,dev,run}
mount --bind /proc $ROOTFS/proc
mount --bind /sys $ROOTFS/sys
mount --bind /dev $ROOTFS/dev
mount --bind /run $ROOTFS/run
Creating a directory tree
The RPM system allows you to easily and quickly prepare the directory tree of the future OS:
PKGSURL=http://mirror.centos.org/centos/8/BaseOS/x86_64/os/Packages
rpm --root=$ROOTFS --initdb
rpm --root=$ROOTFS -ivh
$PKGSURL/centos-release-8.1-1.1911.0.8.el8.x86_64.rpm
$PKGSURL/centos-gpg-keys-8.1-1.1911.0.8.el8.noarch.rpm
$PKGSURL/centos-repos-8.1-1.1911.0.8.el8.x86_64.rpm
dnf --installroot=$ROOTFS --nogpgcheck --setopt=install_weak_deps=False
-y install audit authselect basesystem bash biosdevname coreutils
cronie curl dnf dnf-plugins-core dnf-plugin-spacewalk dracut-config-generic
dracut-config-rescue e2fsprogs filesystem firewalld glibc grub2 grubby hostname
initscripts iproute iprutils iputils irqbalance kbd kernel kernel-tools
kexec-tools less linux-firmware lshw lsscsi ncurses network-scripts
openssh-clients openssh-server passwd plymouth policycoreutils prefixdevname
procps-ng rng-tools rootfiles rpm rsyslog selinux-policy-targeted setup
shadow-utils sssd-kcm sudo systemd util-linux vim-minimal xfsprogs
chrony cloud-init
I consider the last command to be optimally performed in this way, installing specific packages, and be sure to ignore the recommended packages.
If you wish, you can use something like this:
dnf --installroot=$ROOTFS groupinstall base core
--excludepkgs "NetworkManager*"
-e "i*-firmware"
Π yum
no --excludepkgs
, and before you had to install groups, and then remove packages.
The list of packages and dependent groups can be viewed with the command dnf group info core
for the group core
.
OS file customization
Let's create configs for network, fstab, grub2 and use AWS internal 169.254 addresses for DNS and NTP.
cat > $ROOTFS/etc/resolv.conf << HABR
nameserver 169.254.169.253
HABR
cat > $ROOTFS/etc/sysconfig/network << HABR
NETWORKING=yes
NOZEROCONF=yes
HABR
cat > $ROOTFS/etc/sysconfig/network-scripts/ifcfg-eth0 << HABR
DEVICE=eth0
ONBOOT=yes
BOOTPROTO=dhcp
HABR
cat > $ROOTFS/etc/fstab << HABR
LABEL=root / xfs defaults,relatime 1 1
HABR
sed -i "s/cloud-user/centos/" $ROOTFS/etc/cloud/cloud.cfg
echo "server 169.254.169.123 prefer iburst minpoll 4 maxpoll 4" >> $ROOTFS/etc/chrony.conf
sed -i "/^pool /d" $ROOTFS/etc/chrony.conf
sed -i "s/^AcceptEnv/# /" $ROOTFS/etc/ssh/sshd_config
cat > $ROOTFS/etc/default/grub << HABR
GRUB_TIMEOUT=1
GRUB_DISTRIBUTOR="$(sed 's, release .*$,,g' /etc/system-release)"
GRUB_DEFAULT=saved
GRUB_DISABLE_SUBMENU=true
GRUB_TERMINAL_OUTPUT="console"
GRUB_CMDLINE_LINUX="crashkernel=auto console=ttyS0,115200n8 console=tty0 net.ifnames=0 biosdevname=0"
GRUB_DISABLE_RECOVERY="true"
GRUB_ENABLE_BLSCFG=true
HABR
It is here, in GRUB_CMDLINE_LINUX, that I recommend specifying selinux=0, for those who are still afraid of SELinux.
Rebuilding initramfs in chroot
After editing the grub and fstab files, you need to rebuild.
Performing an update:
KERNEL=$(ls $ROOTFS/lib/modules/)
chroot $ROOTFS dracut -f -v /boot/initramfs-$KERNEL.img $KERNEL
chroot $ROOTFS grub2-mkconfig -o /boot/grub2/grub.cfg
chroot $ROOTFS grub2-install $DEVICE
chroot $ROOTFS update-crypto-policies --set FUTURE
Here update-crypto-policies
- optional, for the paranoid π
For "sell", you can do this:
chroot $ROOTFS fips-mode-setup --enable
chroot $ROOTFS grub2-mkconfig -o /boot/grub2/grub.cfg
chroot $ROOTFS grub2-install $DEVICE
After booting the OS, the command update-crypto-policies --show
will issue FIPS.
Autorun and Garbage Cleanup
chroot $ROOTFS systemctl enable network.service
chroot $ROOTFS systemctl enable sshd.service
chroot $ROOTFS systemctl enable cloud-init.service
chroot $ROOTFS systemctl mask tmp.mount
dnf --installroot=$ROOTFS clean all
truncate -c -s 0 $ROOTFS/var/log/*.log
rm -rf var/lib/dnf/*
touch $ROOTFS/.autorelabel
autorelabel
- Needed to automatically install the SELinux file context on first boot.
Now let's unmount the drive:
sync
umount $ROOTFS/{proc,sys,dev,run}
umount $ROOTFS
AMI Registration
To get ami from an ebs disk, you must first take a snapshot of the disk:
aws ec2 create-snapshot
--volume-id vol-09f26eba4c50da110 --region us-east-1
--description 'centos-release-8.1-1.1911.0.8 4.18.0-147.5.1 01'
Some time will have to wait. Let's check the status by the received SnapshotId:
aws ec2 describe-snapshots --region us-east-1 --snapshot-ids snap-0b665542fc59e58ed
When we get "State": "completed"
, you can register an AMI and make it public:
aws ec2 register-image
--region us-east-1
--name 'CentOS-8.1-1.1911.0.8-minimal'
--description 'centos-release-8.1-1.1911.0.8 4.18.0-147.5.1 01'
--virtualization-type hvm --root-device-name /dev/sda1
--block-device-mappings '[{"DeviceName":"/dev/sda1","Ebs": { "SnapshotId": "snap-0b665542fc59e58ed", "VolumeSize":4, "DeleteOnTermination": true, "VolumeType": "gp2"}}]'
--architecture x86_64 --sriov-net-support simple --ena-support
aws ec2 modify-image-attribute
--region us-east-1
--image-id ami-011ed2a37dc89e206
--launch-permission 'Add=[{Group=all}]'
That's all. Now you can launch instances.
In this way, you can make an image, most likely with any Linux distribution. At least for sure Debian (using debootstrap to install a clean system) and the RHEL family.
UPDATED At the request of readers. This process can be automated Packer, only Automate. Here an example template is provided.
Source: habr.com