TL;DR: A comprehensive guide article comparing environments for running applications in containers. It will examine the capabilities of Docker and other similar systems.
A bit of history on where it all started
History
The first widely known method of application isolation is chroot. The corresponding system call changes the root directory, thereby providing the program that invoked it access only to the files within that directory. However, if the program inside is granted superuser rights, it could potentially 'escape' from chroot and gain access to the main operating system. Additionally, other resources (such as RAM and CPU) and network access are not restricted by merely changing the root directory.
The next method is running a full operating system within a container, leveraging mechanisms from the operating system's kernel. Different operating systems refer to this method in various ways, but the essence remains the same — executing several independent operating systems, each operating with the same kernel as the main operating system. This includes FreeBSD Jails, Solaris Zones, OpenVZ, and LXC for Linux. Isolation is ensured not only by disk space but also by other resources, in particular, each container may have limits on CPU time, RAM, and network bandwidth. Compared to chroot, escaping from a container is more challenging since the superuser in the container only has access to the container's internals. However, due to the necessity of maintaining the operating system within the container up to date and the use of older kernel versions (this is particularly relevant for Linux, and less so for FreeBSD), there is a non-zero probability of breaching the kernel's isolation system and gaining access to the main operating system.
Instead of launching a full operating system in a container (with an initialization system, package manager, etc.), applications can be launched directly, as long as they have the necessary libraries and other files available. This idea laid the groundwork for application container virtualization, the most prominent and widely known representative of which is Docker. Compared to previous systems, more flexible isolation mechanisms, along with built-in support for virtual networks between containers and monitoring the application's state within the container, ultimately enable the creation of a unified environment from numerous physical servers for running containers — without the need for manual resource management.
Docker
Docker is the most well-known application containerization software. Written in Go, it utilizes the built-in capabilities of the Linux kernel — cgroups, namespaces, capabilities, etc., as well as file systems like Aufs and others to save disk space.
Source: wikimedia
Architecture
Until version 1.11, Docker operated as a single service that performed all operations with containers: downloading images for containers, launching containers, and handling API requests. Starting from version 1.11, Docker was split into several components that interact with each other: containerd, which manages the entire lifecycle of containers (allocating disk space, downloading images, working with networks, launching, installing, and monitoring the status of containers), and runC, the container runtime based on cgroups and other Linux kernel capabilities. The Docker service itself remains, but it now only serves to handle API requests transmitted to containerd.
Installation and setup
My favorite way to install Docker is using docker-machine, which, in addition to directly installing and configuring Docker on remote servers (including various clouds), allows for interaction with the file systems of remote servers and can also execute various commands.
However, since 2018, the project has seen little development, so we will perform the installation using the standard method for most Linux distributions — by adding the repository and installing the necessary packages.
This method is also applied in automated installations, for example using Ansible or similar systems, but I will not discuss it in this article.
The installation will be performed on CentOS 7, and I will use a virtual machine as a server. To install, simply execute the commands below:
# yum install -y yum-utils
# yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
# yum install docker-ce docker-ce-cli containerd.ioAfter installation, you need to start the service and enable it to start on boot:
# systemctl enable docker
# systemctl start docker
# firewall-cmd --zone=public --add-port=2377/tcp --permanentAdditionally, you can create a docker group, allowing users to work with docker without sudo, set up logging, enable external API access, and remember to configure the firewall more precisely (deny everything that is not allowed; I omitted this in the examples above and below for simplicity and clarity), but I won't go into more detail here.
Other capabilities
In addition to the aforementioned docker machine, there is also docker registry, a tool for storing images for containers, as well as docker compose — a tool for automating application deployment in containers, using YAML files for building and configuring containers and other related items (such as networks, persistent file systems for data storage).
With it, you can also organize pipelines for CI/CD. Another interesting feature is operating in clustered mode, known as swarm mode (prior to version 1.12, it was referred to as docker swarm), enabling multiple servers to form a unified infrastructure for running containers. It supports a virtual network across all servers, has a built-in load balancer, and supports secrets for containers.
Docker compose YAML files, with minor modifications, can be used for such clusters, fully automating the maintenance of small and medium clusters for various purposes. For large clusters, Kubernetes is preferable, as the maintenance costs for swarm mode can exceed those of Kubernetes. Besides runC, you can also install, for example,
Working with Docker
After installation and configuration, let's try to set up a cluster where we will deploy GitLab and Docker Registry for the development team. I will use three virtual machines as servers, on which I will additionally set up a distributed file system GlusterFS, which I will use as storage for docker volumes, for instance, to run a fault-tolerant version of the docker registry. Key components for launching: Docker Registry, PostgreSQL, Redis, GitLab with GitLab Runner support over Swarm. We will run PostgreSQL with clustering. , therefore, GlusterFS should not be used for storing PostgreSQL data. Other critical data will be stored on GlusterFS.
To deploy GlusterFS on all servers (named node1, node2, node3), you need to install packages, allow firewall operations, and create the necessary directories:
# yum -y install centos-release-gluster7
# yum -y install glusterfs-server
# systemctl enable glusterd
# systemctl start glusterd
# firewall-cmd --add-service=glusterfs --permanent
# firewall-cmd --reload
# mkdir -p /srv/gluster
# mkdir -p /srv/docker
# echo "$(hostname):/docker /srv/docker glusterfs defaults,_netdev 0 0" >> /etc/fstabAfter installation, the work of configuring GlusterFS should continue from one node, for example, node1:
# gluster peer probe node2
# gluster peer probe node3
# gluster volume create docker replica 3 node1:/srv/gluster node2:/srv/gluster node3:/srv/gluster force
# gluster volume start dockerThen you need to mount the obtained volume (the command must be executed on all servers):
# mount /srv/dockerSwarm mode configuration is performed on one of the servers, which will be the Leader; the others must join the cluster, so the result of executing the command on the first server must be copied and executed on the others.
Initial cluster setup, command executed on node1:
# docker swarm init
Swarm initialized: current node (a5jpfrh5uvo7svzz1ajduokyq) is now a manager.
To add a worker to this swarm, run the following command:
docker swarm join --token SWMTKN-1-0c5mf7mvzc7o7vjk0wngno2dy70xs95tovfxbv4tqt9280toku-863hyosdlzvd76trfptd4xnzd xx.xx.xx.xx:2377
To add a manager to this swarm, run 'docker swarm join-token manager' and follow the instructions.
# docker swarm join-token managerCopy the result of the second command, execute it on node2 and node3:
# docker swarm join --token SWMTKN-x-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx-xxxxxxxxx xx.xx.xx.xx:2377
This node joined a swarm as a manager.At this point, the preliminary configuration of the servers is complete, we proceed to service configuration, commands will be executed from node1 unless specified otherwise.
First, let's create networks for the containers:
# docker network create --driver=overlay etcd
# docker network create --driver=overlay pgsql
# docker network create --driver=overlay redis
# docker network create --driver=overlay traefik
# docker network create --driver=overlay gitlabNext, we tag the servers; this is necessary for binding certain services to servers:
# docker node update --label-add nodename=node1 node1
# docker node update --label-add nodename=node2 node2
# docker node update --label-add nodename=node3 node3Next, we create directories for storing etcd data, a KV-storage needed for Traefik and Stolon. Similarly to PostgreSQL, these will be containers bound to the servers, so we execute this command on all servers:
# mkdir -p /srv/etcdNext, we create a configuration file for etcd and apply it:
00etcd.yml
version: '3.7'
services:
etcd1:
image: quay.io/coreos/etcd:latest
hostname: etcd1
command:
- etcd
- --name=etcd1
- --data-dir=/data.etcd
- --advertise-client-urls=http://etcd1:2379
- --listen-client-urls=http://0.0.0.0:2379
- --initial-advertise-peer-urls=http://etcd1:2380
- --listen-peer-urls=http://0.0.0.0:2380
- --initial-cluster=etcd1=http://etcd1:2380,etcd2=http://etcd2:2380,etcd3=http://etcd3:2380
- --initial-cluster-state=new
- --initial-cluster-token=etcd-cluster
networks:
- etcd
volumes:
- etcd1vol:/data.etcd
deploy:
replicas: 1
placement:
constraints: [node.labels.nodename == node1]
etcd2:
image: quay.io/coreos/etcd:latest
hostname: etcd2
command:
- etcd
- --name=etcd2
- --data-dir=/data.etcd
- --advertise-client-urls=http://etcd2:2379
- --listen-client-urls=http://0.0.0.0:2379
- --initial-advertise-peer-urls=http://etcd2:2380
- --listen-peer-urls=http://0.0.0.0:2380
- --initial-cluster=etcd1=http://etcd1:2380,etcd2=http://etcd2:2380,etcd3=http://etcd3:2380
- --initial-cluster-state=new
- --initial-cluster-token=etcd-cluster
networks:
- etcd
volumes:
- etcd2vol:/data.etcd
deploy:
replicas: 1
placement:
constraints: [node.labels.nodename == node2]
etcd3:
image: quay.io/coreos/etcd:latest
hostname: etcd3
command:
- etcd
- --name=etcd3
- --data-dir=/data.etcd
- --advertise-client-urls=http://etcd3:2379
- --listen-client-urls=http://0.0.0.0:2379
- --initial-advertise-peer-urls=http://etcd3:2380
- --listen-peer-urls=http://0.0.0.0:2380
- --initial-cluster=etcd1=http://etcd1:2380,etcd2=http://etcd2:2380,etcd3=http://etcd3:2380
- --initial-cluster-state=new
- --initial-cluster-token=etcd-cluster
networks:
- etcd
volumes:
- etcd3vol:/data.etcd
deploy:
replicas: 1
placement:
constraints: [node.labels.nodename == node3]
volumes:
etcd1vol:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/etcd"
etcd2vol:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/etcd"
etcd3vol:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/etcd"
networks:
etcd:
external: true# docker stack deploy --compose-file 00etcd.yml etcdAfter a while, we check that the etcd cluster has started successfully:
# docker exec $(docker ps | awk '/etcd/ {print $1}') etcdctl member list
ade526d28b1f92f7: name=etcd1 peerURLs=http://etcd1:2380 clientURLs=http://etcd1:2379 isLeader=false
bd388e7810915853: name=etcd3 peerURLs=http://etcd3:2380 clientURLs=http://etcd3:2379 isLeader=false
d282ac2ce600c1ce: name=etcd2 peerURLs=http://etcd2:2380 clientURLs=http://etcd2:2379 isLeader=true
# docker exec $(docker ps | awk '/etcd/ {print $1}') etcdctl cluster-health
member ade526d28b1f92f7 is healthy: got healthy result from http://etcd1:2379
member bd388e7810915853 is healthy: got healthy result from http://etcd3:2379
member d282ac2ce600c1ce is healthy: got healthy result from http://etcd2:2379
cluster is healthyWe create directories for Postgresql; execute the command on all servers:
# mkdir -p /srv/pgsqlNext, we create a file for Postgresql configuration:
01pgsql.yml
version: '3.7'
services:
pgsentinel:
image: sorintlab/stolon:master-pg10
command:
- gosu
- stolon
- stolon-sentinel
- --cluster-name=stolon-cluster
- --store-backend=etcdv3
- --store-endpoints=http://etcd1:2379,http://etcd2:2379,http://etcd3:2379
- --log-level=debug
networks:
- etcd
- pgsql
deploy:
replicas: 3
update_config:
parallelism: 1
delay: 30s
order: stop-first
failure_action: pause
pgkeeper1:
image: sorintlab/stolon:master-pg10
hostname: pgkeeper1
command:
- gosu
- stolon
- stolon-keeper
- --pg-listen-address=pgkeeper1
- --pg-repl-username=replica
- --uid=pgkeeper1
- --pg-su-username=postgres
- --pg-su-passwordfile=/run/secrets/pgsql
- --pg-repl-passwordfile=/run/secrets/pgsql_repl
- --data-dir=/var/lib/postgresql/data
- --cluster-name=stolon-cluster
- --store-backend=etcdv3
- --store-endpoints=http://etcd1:2379,http://etcd2:2379,http://etcd3:2379
networks:
- etcd
- pgsql
environment:
- PGDATA=/var/lib/postgresql/data
volumes:
- pgkeeper1:/var/lib/postgresql/data
secrets:
- pgsql
- pgsql_repl
deploy:
replicas: 1
placement:
constraints: [node.labels.nodename == node1]
pgkeeper2:
image: sorintlab/stolon:master-pg10
hostname: pgkeeper2
command:
- gosu
- stolon
- stolon-keeper
- --pg-listen-address=pgkeeper2
- --pg-repl-username=replica
- --uid=pgkeeper2
- --pg-su-username=postgres
- --pg-su-passwordfile=/run/secrets/pgsql
- --pg-repl-passwordfile=/run/secrets/pgsql_repl
- --data-dir=/var/lib/postgresql/data
- --cluster-name=stolon-cluster
- --store-backend=etcdv3
- --store-endpoints=http://etcd1:2379,http://etcd2:2379,http://etcd3:2379
networks:
- etcd
- pgsql
environment:
- PGDATA=/var/lib/postgresql/data
volumes:
- pgkeeper2:/var/lib/postgresql/data
secrets:
- pgsql
- pgsql_repl
deploy:
replicas: 1
placement:
constraints: [node.labels.nodename == node2]
pgkeeper3:
image: sorintlab/stolon:master-pg10
hostname: pgkeeper3
command:
- gosu
- stolon
- stolon-keeper
- --pg-listen-address=pgkeeper3
- --pg-repl-username=replica
- --uid=pgkeeper3
- --pg-su-username=postgres
- --pg-su-passwordfile=/run/secrets/pgsql
- --pg-repl-passwordfile=/run/secrets/pgsql_repl
- --data-dir=/var/lib/postgresql/data
- --cluster-name=stolon-cluster
- --store-backend=etcdv3
- --store-endpoints=http://etcd1:2379,http://etcd2:2379,http://etcd3:2379
networks:
- etcd
- pgsql
environment:
- PGDATA=/var/lib/postgresql/data
volumes:
- pgkeeper3:/var/lib/postgresql/data
secrets:
- pgsql
- pgsql_repl
deploy:
replicas: 1
placement:
constraints: [node.labels.nodename == node3]
postgresql:
image: sorintlab/stolon:master-pg10
command: gosu stolon stolon-proxy --listen-address 0.0.0.0 --cluster-name stolon-cluster --store-backend=etcdv3 --store-endpoints http://etcd1:2379,http://etcd2:2379,http://etcd3:2379
networks:
- etcd
- pgsql
deploy:
replicas: 3
update_config:
parallelism: 1
delay: 30s
order: stop-first
failure_action: rollback
volumes:
pgkeeper1:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/pgsql"
pgkeeper2:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/pgsql"
pgkeeper3:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/pgsql"
secrets:
pgsql:
file: "/srv/docker/postgres"
pgsql_repl:
file: "/srv/docker/replica"
networks:
etcd:
external: true
pgsql:
external: trueGenerating secrets, applying the file:
# </dev/urandom tr -dc 234567890qwertyuopasdfghjkzxcvbnmQWERTYUPASDFGHKLZXCVBNM | head -c $(((RANDOM%3)+15)) > /srv/docker/replica
# </dev/urandom tr -dc 234567890qwertyuopasdfghjkzxcvbnmQWERTYUPASDFGHKLZXCVBNM | head -c $(((RANDOM%3)+15)) > /srv/docker/postgres
# docker stack deploy --compose-file 01pgsql.yml pgsqlAfter some time (checking the output of the command docker service ls, which shows that all services have been started) we initialize the Postgresql cluster:
# docker exec $(docker ps | awk '/pgkeeper/ {print $1}') stolonctl --cluster-name=stolon-cluster --store-backend=etcdv3 --store-endpoints=http://etcd1:2379,http://etcd2:2379,http://etcd3:2379 initChecking the readiness of the Postgresql cluster:
# docker exec $(docker ps | awk '/pgkeeper/ {print $1}') stolonctl --cluster-name=stolon-cluster --store-backend=etcdv3 --store-endpoints=http://etcd1:2379,http://etcd2:2379,http://etcd3:2379 status
=== Active sentinels ===
ID LEADER
26baa11d false
74e98768 false
a8cb002b true
=== Active proxies ===
ID
4d233826
9f562f3b
b0c79ff1
=== Keepers ===
UID HEALTHY PG LISTENADDRESS PG HEALTHY PG WANTEDGENERATION PG CURRENTGENERATION
pgkeeper1 true pgkeeper1:5432 true 2 2
pgkeeper2 true pgkeeper2:5432 true 2 2
pgkeeper3 true pgkeeper3:5432 true 3 3
=== Cluster Info ===
Master Keeper: pgkeeper3
===== Keepers/DB tree =====
pgkeeper3 (master)
├─pgkeeper2
└─pgkeeper1
Configuring traefik to allow external access to the containers:
03traefik.yml
version: '3.7'
services:
traefik:
image: traefik:latest
command: >
--log.level=INFO
--providers.docker=true
--entryPoints.web.address=:80
--providers.providersThrottleDuration=2
--providers.docker.watch=true
--providers.docker.swarmMode=true
--providers.docker.swarmModeRefreshSeconds=15s
--providers.docker.exposedbydefault=false
--accessLog.bufferingSize=0
--api=true
--api.dashboard=true
--api.insecure=true
networks:
- traefik
ports:
- 80:80
volumes:
- /var/run/docker.sock:/var/run/docker.sock
deploy:
replicas: 3
placement:
constraints:
- node.role == manager
preferences:
- spread: node.id
labels:
- traefik.enable=true
- traefik.http.routers.traefik.rule=Host(`traefik.example.com`)
- traefik.http.services.traefik.loadbalancer.server.port=8080
- traefik.docker.network=traefik
networks:
traefik:
external: true# docker stack deploy --compose-file 03traefik.yml traefikStarting Redis Cluster, for this we create a storage directory on all nodes:
# mkdir -p /srv/redis05redis.yml
version: '3.7'
services:
redis-master:
image: 'bitnami/redis:latest'
networks:
- redis
ports:
- '6379:6379'
environment:
- REDIS_REPLICATION_MODE=master
- REDIS_PASSWORD=xxxxxxxxxxx
deploy:
mode: global
restart_policy:
condition: any
volumes:
- 'redis:/opt/bitnami/redis/etc/'
redis-replica:
image: 'bitnami/redis:latest'
networks:
- redis
ports:
- '6379'
depends_on:
- redis-master
environment:
- REDIS_REPLICATION_MODE=slave
- REDIS_MASTER_HOST=redis-master
- REDIS_MASTER_PORT_NUMBER=6379
- REDIS_MASTER_PASSWORD=xxxxxxxxxxx
- REDIS_PASSWORD=xxxxxxxxxxx
deploy:
mode: replicated
replicas: 3
update_config:
parallelism: 1
delay: 10s
restart_policy:
condition: any
redis-sentinel:
image: 'bitnami/redis:latest'
networks:
- redis
ports:
- '16379'
depends_on:
- redis-master
- redis-replica
entrypoint: |
bash -c 'bash -s <<EOF
"/bin/bash" -c "cat < /opt/bitnami/redis/etc/sentinel.conf
port 16379
dir /tmp
sentinel monitor master-node redis-master 6379 2
sentinel down-after-milliseconds master-node 5000
sentinel parallel-syncs master-node 1
sentinel failover-timeout master-node 5000
sentinel auth-pass master-node xxxxxxxxxxx
sentinel announce-ip redis-sentinel
sentinel announce-port 16379
EOF"
"/bin/bash" -c "redis-sentinel /opt/bitnami/redis/etc/sentinel.conf"
EOF'
deploy:
mode: global
restart_policy:
condition: any
volumes:
redis:
driver: local
driver_opts:
type: 'none'
o: 'bind'
device: "/srv/redis"
networks:
redis:
external: true# docker stack deploy --compose-file 05redis.yml redisAdding Docker Registry:
06registry.yml
version: '3.7'
services:
registry:
image: registry:2.6
networks:
- traefik
volumes:
- registry_data:/var/lib/registry
deploy:
replicas: 1
placement:
constraints: [node.role == manager]
restart_policy:
condition: on-failure
labels:
- traefik.enable=true
- traefik.http.routers.registry.rule=Host(`registry.example.com`)
- traefik.http.services.registry.loadbalancer.server.port=5000
- traefik.docker.network=traefik
volumes:
registry_data:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/docker/registry"
networks:
traefik:
external: true# mkdir /srv/docker/registry
# docker stack deploy --compose-file 06registry.yml registryAnd finally — GitLab:
08gitlab-runner.yml
version: '3.7'
services:
gitlab:
image: gitlab/gitlab-ce:latest
networks:
- pgsql
- redis
- traefik
- gitlab
ports:
- 22222:22
environment:
GITLAB_OMNIBUS_CONFIG: |
postgresql['enable'] = false
redis['enable'] = false
gitlab_rails['registry_enabled'] = false
gitlab_rails['db_username'] = "gitlab"
gitlab_rails['db_password'] = "XXXXXXXXXXX"
gitlab_rails['db_host'] = "postgresql"
gitlab_rails['db_port'] = "5432"
gitlab_rails['db_database'] = "gitlab"
gitlab_rails['db_adapter'] = 'postgresql'
gitlab_rails['db_encoding'] = 'utf8'
gitlab_rails['redis_host'] = 'redis-master'
gitlab_rails['redis_port'] = '6379'
gitlab_rails['redis_password'] = 'xxxxxxxxxxx'
gitlab_rails['smtp_enable'] = true
gitlab_rails['smtp_address'] = "smtp.yandex.ru"
gitlab_rails['smtp_port'] = 465
gitlab_rails['smtp_user_name'] = "noreply@example.com"
gitlab_rails['smtp_password'] = "xxxxxxxxx"
gitlab_rails['smtp_domain'] = "example.com"
gitlab_rails['gitlab_email_from'] = 'noreply@example.com'
gitlab_rails['smtp_authentication'] = "login"
gitlab_rails['smtp_tls'] = true
gitlab_rails['smtp_enable_starttls_auto'] = true
gitlab_rails['smtp_openssl_verify_mode'] = 'peer'
external_url 'http://gitlab.example.com/'
gitlab_rails['gitlab_shell_ssh_port'] = 22222
volumes:
- gitlab_conf:/etc/gitlab
- gitlab_logs:/var/log/gitlab
- gitlab_data:/var/opt/gitlab
deploy:
mode: replicated
replicas: 1
placement:
constraints:
- node.role == manager
labels:
- traefik.enable=true
- traefik.http.routers.gitlab.rule=Host(`gitlab.example.com`)
- traefik.http.services.gitlab.loadbalancer.server.port=80
- traefik.docker.network=traefik
gitlab-runner:
image: gitlab/gitlab-runner:latest
networks:
- gitlab
volumes:
- gitlab_runner_conf:/etc/gitlab
- /var/run/docker.sock:/var/run/docker.sock
deploy:
mode: replicated
replicas: 1
placement:
constraints:
- node.role == manager
volumes:
gitlab_conf:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/docker/gitlab/conf"
gitlab_logs:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/docker/gitlab/logs"
gitlab_data:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/docker/gitlab/data"
gitlab_runner_conf:
driver: local
driver_opts:
type: none
o: bind
device: "/srv/docker/gitlab/runner"
networks:
pgsql:
external: true
redis:
external: true
traefik:
external: true
gitlab:
external: true# mkdir -p /srv/docker/gitlab/conf
# mkdir -p /srv/docker/gitlab/logs
# mkdir -p /srv/docker/gitlab/data
# mkdir -p /srv/docker/gitlab/runner
# docker stack deploy --compose-file 08gitlab-runner.yml gitlabFinal state of the cluster and services:
# docker service ls
ID NAME MODE REPLICAS IMAGE PORTS
lef9n3m92buq etcd_etcd1 replicated 1/1 quay.io/coreos/etcd:latest
ij6uyyo792x5 etcd_etcd2 replicated 1/1 quay.io/coreos/etcd:latest
fqttqpjgp6pp etcd_etcd3 replicated 1/1 quay.io/coreos/etcd:latest
hq5iyga28w33 gitlab_gitlab replicated 1/1 gitlab/gitlab-ce:latest *:22222->22/tcp
dt7s6vs0q4qc gitlab_gitlab-runner replicated 1/1 gitlab/gitlab-runner:latest
k7uoezno0h9n pgsql_pgkeeper1 replicated 1/1 sorintlab/stolon:master-pg10
cnrwul4r4nse pgsql_pgkeeper2 replicated 1/1 sorintlab/stolon:master-pg10
frflfnpty7tr pgsql_pgkeeper3 replicated 1/1 sorintlab/stolon:master-pg10
x7pqqchi52kq pgsql_pgsentinel replicated 3/3 sorintlab/stolon:master-pg10
mwu2wl8fti4r pgsql_postgresql replicated 3/3 sorintlab/stolon:master-pg10
9hkbe2vksbzb redis_redis-master global 3/3 bitnami/redis:latest *:6379->6379/tcp
l88zn8cla7dc redis_redis-replica replicated 3/3 bitnami/redis:latest *:30003->6379/tcp
1utp309xfmsy redis_redis-sentinel global 3/3 bitnami/redis:latest *:30002->16379/tcp
oteb824ylhyp registry_registry replicated 1/1 registry:2.6
qovrah8nzzu8 traefik_traefik replicated 3/3 traefik:latest *:80->80/tcp, *:443->443/tcpWhat else can be improved? It is essential to configure Traefik for HTTPS operation of containers, add TLS encryption for PostgreSQL and Redis. Overall, it can already be delivered to developers as a PoC. Now let's explore alternatives to Docker.
Podman
Another well-known engine for running containers, grouped by pods (groups of containers deployed together). Unlike Docker, it does not require any service to run containers; all operations are performed through the libpod library. It is also written in Go and requires an OCI-compatible runtime for container execution, such as runC.
Working with Podman generally resembles working with Docker, to the extent that you can set it up like this (as reported by many who have tried, including the author of this article):
$ alias docker=podmanand you can continue working. The situation with Podman is very interesting because while earlier versions of Kubernetes worked with Docker, since around 2015, following the standardization of the container world (OCI — Open Container Initiative) and the division of Docker into containerd and runC, an alternative to Docker for running in Kubernetes has emerged: CRI-O. In this regard, Podman serves as an alternative to Docker, built on Kubernetes principles, including the grouping of containers, but the main aim of the project is to run containers in a Docker-like manner without additional services. For obvious reasons, there is no swarm mode, as developers clearly state that if a cluster is needed, Kubernetes should be used.
Installation
To install on CentOS 7, simply activate the Extras repository, and then install everything with the command:
# yum -y install podmanOther capabilities
Podman can generate units for systemd, thus addressing the task of starting containers after the server reboots. Additionally, correct operation of systemd as pid 1 in the container is claimed. For building containers, there is a separate tool called buildah, and there are also third-party tools — analogues of docker-compose, which can generate configuration files compatible with Kubernetes, simplifying the transition from Podman to Kubernetes as much as possible.
Working with Podman
Since there is no swarm mode (a transition to Kubernetes is assumed if a cluster is needed) — we will be building individual containers.
Let's install podman-compose:
# yum -y install python3-pip
# pip3 install podman-composeThe resulting configuration file for podman is slightly different; for instance, a separate volumes section had to be moved directly into the services section.
gitlab-podman.yml
version: '3.7'
services:
gitlab:
image: gitlab/gitlab-ce:latest
hostname: gitlab.example.com
restart: unless-stopped
environment:
GITLAB_OMNIBUS_CONFIG: |
gitlab_rails['gitlab_shell_ssh_port'] = 22222
ports:
- "80:80"
- "22222:22"
volumes:
- /srv/podman/gitlab/conf:/etc/gitlab
- /srv/podman/gitlab/data:/var/opt/gitlab
- /srv/podman/gitlab/logs:/var/log/gitlab
networks:
- gitlab
gitlab-runner:
image: gitlab/gitlab-runner:alpine
restart: unless-stopped
depends_on:
- gitlab
volumes:
- /srv/podman/gitlab/runner:/etc/gitlab-runner
- /var/run/docker.sock:/var/run/docker.sock
networks:
- gitlab
networks:
gitlab:# podman-compose -f gitlab-runner.yml -d upResult:
# podman ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
da53da946c01 docker.io/gitlab/gitlab-runner:alpine run --user=gitlab... About a minute ago Up About a minute ago 0.0.0.0:22222->22/tcp, 0.0.0.0:80->80/tcp root_gitlab-runner_1
781c0103c94a docker.io/gitlab/gitlab-ce:latest /assets/wrapper About a minute ago Up About a minute ago 0.0.0.0:22222->22/tcp, 0.0.0.0:80->80/tcp root_gitlab_1Let's see what it generates for systemd and Kubernetes; for that, we need to know the name or ID of the pod:
# podman pod ls
POD ID NAME STATUS CREATED # OF CONTAINERS INFRA ID
71fc2b2a5c63 root Running 11 minutes ago 3 db40ab8bf84bKubernetes:
# podman generate kube 71fc2b2a5c63
# Generation of Kubernetes YAML is still under development!
#
# Save the output of this file and use kubectl create -f to import
# it into Kubernetes.
#
# Created with podman-1.6.4
apiVersion: v1
kind: Pod
metadata:
creationTimestamp: "2020-07-29T19:22:40Z"
labels:
app: root
name: root
spec:
containers:
- command:
- /assets/wrapper
env:
- name: PATH
value: /opt/gitlab/embedded/bin:/opt/gitlab/bin:/assets:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
- name: TERM
value: xterm
- name: HOSTNAME
value: gitlab.example.com
- name: container
value: podman
- name: GITLAB_OMNIBUS_CONFIG
value: |
gitlab_rails['gitlab_shell_ssh_port'] = 22222
- name: LANG
value: C.UTF-8
image: docker.io/gitlab/gitlab-ce:latest
name: rootgitlab1
ports:
- containerPort: 22
hostPort: 22222
protocol: TCP
- containerPort: 80
hostPort: 80
protocol: TCP
resources: {}
securityContext:
allowPrivilegeEscalation: true
capabilities: {}
privileged: false
readOnlyRootFilesystem: false
volumeMounts:
- mountPath: /var/opt/gitlab
name: srv-podman-gitlab-data
- mountPath: /var/log/gitlab
name: srv-podman-gitlab-logs
- mountPath: /etc/gitlab
name: srv-podman-gitlab-conf
workingDir: /
- command:
- run
- --user=gitlab-runner
- --working-directory=/home/gitlab-runner
env:
- name: PATH
value: /usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
- name: TERM
value: xterm
- name: HOSTNAME
- name: container
value: podman
image: docker.io/gitlab/gitlab-runner:alpine
name: rootgitlab-runner1
resources: {}
securityContext:
allowPrivilegeEscalation: true
capabilities: {}
privileged: false
readOnlyRootFilesystem: false
volumeMounts:
- mountPath: /etc/gitlab-runner
name: srv-podman-gitlab-runner
- mountPath: /var/run/docker.sock
name: var-run-docker.sock
workingDir: /
volumes:
- hostPath:
path: /srv/podman/gitlab/runner
type: Directory
name: srv-podman-gitlab-runner
- hostPath:
path: /var/run/docker.sock
type: File
name: var-run-docker.sock
- hostPath:
path: /srv/podman/gitlab/data
type: Directory
name: srv-podman-gitlab-data
- hostPath:
path: /srv/podman/gitlab/logs
type: Directory
name: srv-podman-gitlab-logs
- hostPath:
path: /srv/podman/gitlab/conf
type: Directory
name: srv-podman-gitlab-conf
status: {}Systemd:
# podman generate systemd 71fc2b2a5c63
# pod-71fc2b2a5c6346f0c1c86a2dc45dbe78fa192ea02aac001eb8347ccb8c043c26.service
# autogenerated by Podman 1.6.4
# Thu Jul 29 15:23:28 EDT 2020
[Unit]
Description=Podman pod-71fc2b2a5c6346f0c1c86a2dc45dbe78fa192ea02aac001eb8347ccb8c043c26.service
Documentation=man:podman-generate-systemd(1)
Requires=container-781c0103c94aaa113c17c58d05ddabf8df4bf39707b664abcf17ed2ceff467d3.service container-da53da946c01449f500aa5296d9ea6376f751948b17ca164df438b7df6607864.service
Before=container-781c0103c94aaa113c17c58d05ddabf8df4bf39707b664abcf17ed2ceff467d3.service container-da53da946c01449f500aa5296d9ea6376f751948b17ca164df438b7df6607864.service
[Service]
Restart=on-failure
ExecStart=/usr/bin/podman start db40ab8bf84bf35141159c26cb6e256b889c7a98c0418eee3c4aa683c14fccaa
ExecStop=/usr/bin/podman stop -t 10 db40ab8bf84bf35141159c26cb6e256b889c7a98c0418eee3c4aa683c14fccaa
KillMode=none
Type=forking
PIDFile=/var/run/containers/storage/overlay-containers/db40ab8bf84bf35141159c26cb6e256b889c7a98c0418eee3c4aa683c14fccaa/userdata/conmon.pid
[Install]
WantedBy=multi-user.target
# container-da53da946c01449f500aa5296d9ea6376f751948b17ca164df438b7df6607864.service
# autogenerated by Podman 1.6.4
# Thu Jul 29 15:23:28 EDT 2020
[Unit]
Description=Podman container-da53da946c01449f500aa5296d9ea6376f751948b17ca164df438b7df6607864.service
Documentation=man:podman-generate-systemd(1)
RefuseManualStart=yes
RefuseManualStop=yes
BindsTo=pod-71fc2b2a5c6346f0c1c86a2dc45dbe78fa192ea02aac001eb8347ccb8c043c26.service
After=pod-71fc2b2a5c6346f0c1c86a2dc45dbe78fa192ea02aac001eb8347ccb8c043c26.service
[Service]
Restart=on-failure
ExecStart=/usr/bin/podman start da53da946c01449f500aa5296d9ea6376f751948b17ca164df438b7df6607864
ExecStop=/usr/bin/podman stop -t 10 da53da946c01449f500aa5296d9ea6376f751948b17ca164df438b7df6607864
KillMode=none
Type=forking
PIDFile=/var/run/containers/storage/overlay-containers/da53da946c01449f500aa5296d9ea6376f751948b17ca164df438b7df6607864/userdata/conmon.pid
[Install]
WantedBy=multi-user.target
# container-781c0103c94aaa113c17c58d05ddabf8df4bf39707b664abcf17ed2ceff467d3.service
# autogenerated by Podman 1.6.4
# Thu Jul 29 15:23:28 EDT 2020
[Unit]
Description=Podman container-781c0103c94aaa113c17c58d05ddabf8df4bf39707b664abcf17ed2ceff467d3.service
Documentation=man:podman-generate-systemd(1)
RefuseManualStart=yes
RefuseManualStop=yes
BindsTo=pod-71fc2b2a5c6346f0c1c86a2dc45dbe78fa192ea02aac001eb8347ccb8c043c26.service
After=pod-71fc2b2a5c6346f0c1c86a2dc45dbe78fa192ea02aac001eb8347ccb8c043c26.service
[Service]
Restart=on-failure
ExecStart=/usr/bin/podman start 781c0103c94aaa113c17c58d05ddabf8df4bf39707b664abcf17ed2ceff467d3
ExecStop=/usr/bin/podman stop -t 10 781c0103c94aaa113c17c58d05ddabf8df4bf39707b664abcf17ed2ceff467d3
KillMode=none
Type=forking
PIDFile=/var/run/containers/storage/overlay-containers/781c0103c94aaa113c17c58d05ddabf8df4bf39707b664abcf17ed2ceff467d3/userdata/conmon.pid
[Install]
WantedBy=multi-user.targetUnfortunately, apart from starting containers, the generated unit for systemd does nothing else (for example, it does not clean up old containers when restarting such a service), so you will have to add those things manually.
In principle, Podman is sufficient to try what containers are, to migrate old Docker Compose configurations, and then switch to Kubernetes if you need a cluster or get a simpler alternative to Docker.
rkt
Project about six months ago because it was acquired by RedHat, so I won’t dwell on it in detail. Overall, it left a rather good impression, but compared to Docker and especially Podman, it seems like a complex tool. There was also a CoreOS distribution built on the basis of rkt (although they originally had Docker), but its support also ended after the RedHat acquisition.
Plash
More , the author of which simply wanted to collect and launch containers. Judging by the documentation and code — the author did not adhere to standards but simply decided to write their own implementation, which is basically what they did.
Conclusions
The situation with Kubernetes is quite interesting: on the one hand, with Docker, you can build a cluster (in swarm mode), which can even run production environments for clients. This is particularly relevant for small teams (3-5 people) or under low overall load, or when there is no desire to delve into the intricacies of configuring Kubernetes, including for high loads.
Podman does not offer full compatibility, but it does have one important advantage — compatibility with Kubernetes, including additional tools (buildah and others). Therefore, I will approach the choice of the tool as follows: for small teams or with a limited budget — Docker (with possible swarm mode), for personal development on a local localhost — Podman with friends, and for everyone else — Kubernetes.
I'm not sure that the situation with Docker won't change in the future, after all they are pioneers and are gradually getting standardized, but Podman, with all its drawbacks (only working on Linux, no clustering, builds and other tasks — handled by third-party solutions), has a clearer future. Therefore, I invite everyone interested to discuss these conclusions in the comments.
P.S. On August 3, we are launching "", where you can learn more about its operation. We will break down all its tools: from basic abstractions to network parameters, nuances of working with different OS and programming languages. You will get acquainted with the technology and understand where and how to best use Docker. We will also share best practice case studies.
Pre-order price before release: 5000 RUB. You can learn about the "Docker Video Course" program .
Source: habr.com
