Let's create a minimal Docker image which contains an arbitrary Go application. As an example application, we can use "Hello world":
cmd/main.go:
package main
cmd/main.go:
package main
import (
"fmt"
)
"fmt"
)
func main() {
fmt.Println("Hello, world!")
}
To build it and place the executable in bin directory we need to run go build:
$ go build -o bin/hello-world cmd/main.go
To test the executable, let's run it:
$ ./bin/hello-world
Hello, world!
We want to create a Docker image which, when started, runs this binary. We first have to add Dockerfile - a file which defines how will Docker image be created. Creating a Docker image is like creating a lasagne: we take a base layer and then keep adding new layers on top of each other. Dockerfile specifies what will be the base Docker image (base layer), which application has to reside in it, what is its environment and dependencies that have to be installed and also how will that app be executed (or, what shall be executed when that image is launched).
In our case, we only want to have a single binary in the container and we want it to be launched. For this use case, our Dockerfile can be like this:
go-docker-hello-world/Dockerfile:
FROM scratch
COPY bin/hello-world app/
CMD ["/app/hello-world"]
FROM scratch specifies that empty image (0 bytes!) shall be used as a base layer (or...we can say that there is no base layer).
COPY copies files or directories from source in the host to destination in the container. Working directory on host can be specified via context argument to docker build command. Current directory is used by default and in our case that's go-docker-hello-world. Our binary will be copied here from bin directory on host into the app directory in the container. If destination has to be directory, a slash (/) hast to be added after the destination name. If we didn't add slash, COPY would have copied our binary into the root directory of the container and would have renamed it to app.
CMD contains the name of the executable that has to be run upon container's launch. We need to use an array format (square brackets) as Docker then uses the first argument as the entry point (process that is executed first) and subsequent elements are its arguments. If we used "/app/hello-world" instead of ["/app/hello-world"] Docker would have tried to pass the name of the executable as an argument to /bin/sh but as base image is empty, shell is not present and we'd get an error when running the container:
go-docker-hello-world$ docker build -t helloworld ./
Sending build context to Docker daemon 2.055MB
Step 1/3 : FROM scratch
--->
Step 2/3 : COPY bin/hello-world app/
---> 7c0c34e6ad64
Step 3/3 : CMD ["/app/hello-world"]
---> Running in b3f5695b79c5
Removing intermediate container b3f5695b79c5
---> 171dbd862be1
Successfully built 171dbd862be1
Successfully tagged helloworld:latest
-t applies a tag (name) to the container which can be used later in container managing commands (it is easier to use some descriptive name rather than container ID which is just an array of numbers).
./ specifies the context (the current working directory) for commands in the Dockerfile.
Let's verify that it appears in the list of images:
$ docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
helloworld latest 171dbd862be1 42 minutes ago 2MB
Let's inspect it to verify that entry point is indeed our application:
$ docker inspect 171dbd862be1
[
{
"Id": "sha256:171dbd862be107306bcad870587f8961c00566b946a4d2717ccbf3863492ca2c",
"RepoTags": [
"helloworld:latest"
],
"RepoDigests": [],
"Parent": "sha256:7c0c34e6ad64538ff493910efd6046043b6fa28e78015be6333fcd2e880122d4",
"Comment": "",
"Created": "2019-03-22T16:15:30.7049579Z",
"Container": "b3f5695b79c5add5e86af2ea02b893bd5ed35381221cca1fbf84dd6ea401b69e",
"ContainerConfig": {
"Hostname": "b3f5695b79c5",
"Domainname": "",
"User": "",
"AttachStdin": false,
"AttachStdout": false,
"AttachStderr": false,
"Tty": false,
"OpenStdin": false,
"StdinOnce": false,
"Env": [
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
],
"Cmd": [
"/bin/sh",
"-c",
"#(nop) ",
"CMD [\"/app/hello-world\"]"
],
"ArgsEscaped": true,
"Image": "sha256:7c0c34e6ad64538ff493910efd6046043b6fa28e78015be6333fcd2e880122d4",
"Volumes": null,
"WorkingDir": "",
"Entrypoint": null,
"OnBuild": null,
"Labels": {}
},
"DockerVersion": "18.09.3",
"Author": "",
"Config": {
"Hostname": "",
"Domainname": "",
"User": "",
"AttachStdin": false,
"AttachStdout": false,
"AttachStderr": false,
"Tty": false,
"OpenStdin": false,
"StdinOnce": false,
"Env": [
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
],
"Cmd": [
"/app/hello-world"
],
"ArgsEscaped": true,
"Image": "sha256:7c0c34e6ad64538ff493910efd6046043b6fa28e78015be6333fcd2e880122d4",
"Volumes": null,
"WorkingDir": "",
"Entrypoint": null,
"OnBuild": null,
"Labels": null
},
"Architecture": "amd64",
"Os": "linux",
"Size": 1997502,
"VirtualSize": 1997502,
"GraphDriver": {
"Data": {
"MergedDir": "/var/lib/docker/overlay2/86d02e448ac1c650f65d6eb30b21eeea2f13f176918ccd6af3440c0d89336b19/merged",
"UpperDir": "/var/lib/docker/overlay2/86d02e448ac1c650f65d6eb30b21eeea2f13f176918ccd6af3440c0d89336b19/diff",
"WorkDir": "/var/lib/docker/overlay2/86d02e448ac1c650f65d6eb30b21eeea2f13f176918ccd6af3440c0d89336b19/work"
},
"Name": "overlay2"
},
"RootFS": {
"Type": "layers",
"Layers": [
"sha256:60fdb797c60194a24fa8135f6a1dbe2ed03172037ff5e63eedfc372c2a92964d"
]
},
"Metadata": {
"LastTagTime": "2019-03-22T16:15:30.835954727Z"
}
}
]
Finally, let's run the container:
$ docker run helloworld
Hello, world!
When I built once natively, on Ubuntu, a similar, small app from scratch, I got the following error when I ran its container:
To build it and place the executable in bin directory we need to run go build:
$ go build -o bin/hello-world cmd/main.go
To test the executable, let's run it:
$ ./bin/hello-world
Hello, world!
We want to create a Docker image which, when started, runs this binary. We first have to add Dockerfile - a file which defines how will Docker image be created. Creating a Docker image is like creating a lasagne: we take a base layer and then keep adding new layers on top of each other. Dockerfile specifies what will be the base Docker image (base layer), which application has to reside in it, what is its environment and dependencies that have to be installed and also how will that app be executed (or, what shall be executed when that image is launched).
In our case, we only want to have a single binary in the container and we want it to be launched. For this use case, our Dockerfile can be like this:
go-docker-hello-world/Dockerfile:
FROM scratch
COPY bin/hello-world app/
CMD ["/app/hello-world"]
FROM scratch specifies that empty image (0 bytes!) shall be used as a base layer (or...we can say that there is no base layer).
COPY copies files or directories from source in the host to destination in the container. Working directory on host can be specified via context argument to docker build command. Current directory is used by default and in our case that's go-docker-hello-world. Our binary will be copied here from bin directory on host into the app directory in the container. If destination has to be directory, a slash (/) hast to be added after the destination name. If we didn't add slash, COPY would have copied our binary into the root directory of the container and would have renamed it to app.
CMD contains the name of the executable that has to be run upon container's launch. We need to use an array format (square brackets) as Docker then uses the first argument as the entry point (process that is executed first) and subsequent elements are its arguments. If we used "/app/hello-world" instead of ["/app/hello-world"] Docker would have tried to pass the name of the executable as an argument to /bin/sh but as base image is empty, shell is not present and we'd get an error when running the container:
docker: Error response from daemon: OCI runtime create failed: container_linux.go:344: starting container process caused "exec: \"/bin/sh\": stat /bin/sh: no such file or directory": unknown.Let's create the container:
go-docker-hello-world$ docker build -t helloworld ./
Sending build context to Docker daemon 2.055MB
Step 1/3 : FROM scratch
--->
Step 2/3 : COPY bin/hello-world app/
---> 7c0c34e6ad64
Step 3/3 : CMD ["/app/hello-world"]
---> Running in b3f5695b79c5
Removing intermediate container b3f5695b79c5
---> 171dbd862be1
Successfully built 171dbd862be1
Successfully tagged helloworld:latest
-t applies a tag (name) to the container which can be used later in container managing commands (it is easier to use some descriptive name rather than container ID which is just an array of numbers).
./ specifies the context (the current working directory) for commands in the Dockerfile.
Let's verify that it appears in the list of images:
$ docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
helloworld latest 171dbd862be1 42 minutes ago 2MB
Let's inspect it to verify that entry point is indeed our application:
$ docker inspect 171dbd862be1
[
{
"Id": "sha256:171dbd862be107306bcad870587f8961c00566b946a4d2717ccbf3863492ca2c",
"RepoTags": [
"helloworld:latest"
],
"RepoDigests": [],
"Parent": "sha256:7c0c34e6ad64538ff493910efd6046043b6fa28e78015be6333fcd2e880122d4",
"Comment": "",
"Created": "2019-03-22T16:15:30.7049579Z",
"Container": "b3f5695b79c5add5e86af2ea02b893bd5ed35381221cca1fbf84dd6ea401b69e",
"ContainerConfig": {
"Hostname": "b3f5695b79c5",
"Domainname": "",
"User": "",
"AttachStdin": false,
"AttachStdout": false,
"AttachStderr": false,
"Tty": false,
"OpenStdin": false,
"StdinOnce": false,
"Env": [
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
],
"Cmd": [
"/bin/sh",
"-c",
"#(nop) ",
"CMD [\"/app/hello-world\"]"
],
"ArgsEscaped": true,
"Image": "sha256:7c0c34e6ad64538ff493910efd6046043b6fa28e78015be6333fcd2e880122d4",
"Volumes": null,
"WorkingDir": "",
"Entrypoint": null,
"OnBuild": null,
"Labels": {}
},
"DockerVersion": "18.09.3",
"Author": "",
"Config": {
"Hostname": "",
"Domainname": "",
"User": "",
"AttachStdin": false,
"AttachStdout": false,
"AttachStderr": false,
"Tty": false,
"OpenStdin": false,
"StdinOnce": false,
"Env": [
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
],
"Cmd": [
"/app/hello-world"
],
"ArgsEscaped": true,
"Image": "sha256:7c0c34e6ad64538ff493910efd6046043b6fa28e78015be6333fcd2e880122d4",
"Volumes": null,
"WorkingDir": "",
"Entrypoint": null,
"OnBuild": null,
"Labels": null
},
"Architecture": "amd64",
"Os": "linux",
"Size": 1997502,
"VirtualSize": 1997502,
"GraphDriver": {
"Data": {
"MergedDir": "/var/lib/docker/overlay2/86d02e448ac1c650f65d6eb30b21eeea2f13f176918ccd6af3440c0d89336b19/merged",
"UpperDir": "/var/lib/docker/overlay2/86d02e448ac1c650f65d6eb30b21eeea2f13f176918ccd6af3440c0d89336b19/diff",
"WorkDir": "/var/lib/docker/overlay2/86d02e448ac1c650f65d6eb30b21eeea2f13f176918ccd6af3440c0d89336b19/work"
},
"Name": "overlay2"
},
"RootFS": {
"Type": "layers",
"Layers": [
"sha256:60fdb797c60194a24fa8135f6a1dbe2ed03172037ff5e63eedfc372c2a92964d"
]
},
"Metadata": {
"LastTagTime": "2019-03-22T16:15:30.835954727Z"
}
}
]
Finally, let's run the container:
$ docker run helloworld
Hello, world!
When I built once natively, on Ubuntu, a similar, small app from scratch, I got the following error when I ran its container:
ERROR: for my_app Cannot start service carl: OCI runtime create failed: container_linux.go:345: starting container process caused "exec: \"my_app\": executable file not found in $PATH": unknown
The problem seemed to be related to app being dynamically linked to some of shared libraries on my dev Linux machine so when binary was copied over to empty (scratch) Docker container, binary could not find them so threw such error. I assume this was the reason for such error as solution was to explicitly disable dynamic linking when building my app:
$ CGO_ENABLED=0 go build cmd/main.go
Indeed, using CGO_ENABLED flag makes a difference.
Default is dynamic linking:
$ go build cmd/main.go
$ file main
main: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/l, not stripped
Static linking has to be explicitly set:
$ CGO_ENABLED=0 go build cmd/main.go
$ file main
main: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped
