Can someone explain to me why the normal Docker process is to build an image from a Dockerfile and then upload it to a repository, instead of just moving the Dockerfile to and from the repository?
Let's say we have a development laptop and a test server with Docker.
If we build the image, that means uploading and downloading all of the packages inside the Dockerfile. Sometimes this can be very large (e.g. PyTorch > 500MB).
Instead of transporting the large imagefile to and from the server, doesn't it make sense to, perhaps compile the image locally to verify it works, but mostly transport the small Dockerfile and build the image on the server?
This started out as a comment, but it got too long. It is likely to not be a comprehensive answer, but may contain useful information regardless.
Often the Dockerfile will form part of a larger build process, with output files from previous stages being copied into the final image. If you want to host the Dockerfile instead of the final image, you’d also have to host either the (usually temporary) processed files or the entire source repo & build script.
The latter is often done for open source projects, but for convenience pre-built Docker images are also frequently available.
One tidy solution to this problem is to write the entire build process in the Dockerfile using multi-stage builds (introduced in Docker CE 17.05 & EE 17.06). But even with the complete build process described in a platform-independent manner in a single Dockerfile, the complete source repository must still be provided.
TL,DR: Think of a Docker image as a regular binary. It’s convenient to download and install without messing around with source files. You could download the source for a C application and build it using the provided Makefile, but why would you if a binary was made available for your system?
Instead of transporting the large imagefile to and from the server,
doesn't it make sense to, perhaps compile the image locally to verify
it works, but mostly transport the small Dockerfile and build the
image on the server?
Absolutely! You can, for example, set up an automated build on Docker Hub which will do just that every time you check in an updated version of your Dockerfile to your GitHub repo.
Or you can set up your own build server / CI pipeline accordingly.
IMHO, one of the reason for building the images concept and putting into repository is sharing with people too. For example we call Python's out of the box image for performing all python related stuff for a python program to run in Dockerfile. Similarly we could create a custom code(let's take example I did for apache installation with some custom steps(like ports changes and additionally doing some steps) I created its image and then finally put it to my company's repository.
I came to know after few days that may other teams are using it too and now when they are sharing it they need NOT to make any changes simply use my image and they should be done.
Related
We use JFrog Artifactory for managing Docker images created from Dockerfiles. It has a nice feature where you can see all the "layers" that were involved in creating any given final Docker image.
We have to be careful though, so that credentials do not wind up showing in the layers where they were used. The way we currently do this is by using multistage builds with "COPY --from".
However recently we needed to use credentials for a particular yum repository, which is supplying many dependencies that we need (thousands of files spread throughout the file system). I used yum-config-manager to set the password and username from ENV variables. However even if I use FROM depbuilder, the commands from all the prior stages (including depbuilder) now become visible in Artifactory.
I need to avoid that from happening, and a colleage suggested that we could simply do this:
COPY --from=depbuilder / /
And that way it wouldn't show the other stage's steps as part of the history of the build in Artifactory. However I'm afraid that this command might not set all the ownership and permissions correctly, or it might miss certain files, since the documentation on how it works seems spotty at best.
So what's the best way to copy everything from a prior build stage in a way that would be invisible to someone looking at the build layers in Artifactory?
Here is a screenshot showing what the layers look like in Artifactory: (if we expand the RUN steps, currently we could see the credentials passed into docker via ENV since they become part of the URL for Artifactory)
Thanks for any help!
Context: Reading through this blog post.
Pushing images to a registry seems to be the "right thing to do" ... but I don't understand why.
What purpose does this serve? Is it because the server I ssh into needs to have a local copy of the image? And to do that, one approach is to pull an image from a registry?
What purpose does this serve? Is it because the server I ssh into needs to have a local copy of the image? And to do that, one approach is to pull an image from a registry?
From the CI/CD perspective, a docker registry is the equivalent of an artifact repository for images. You want a central source of these images to download from as you go from one docker host to another since your build server is most likely different than your dev and prod servers.
Couldn't I just upload an image from one machine (say a CI/CD server) via ssh? using dockerhub seems needlessly ceremonious to me. Like in this example (I know this api is deprecated but it illustrates my point).
It is possible to save/load images directly to a docker host, but there a few major downsides. First, you lose any benefit from docker's layered filesystem. When building an app in CI/CD, most of the time only the last few layers should need to be rebuilt with your application changes. There should be the same previous base image and various common layers to build your app that remain identical. With a registry, these common layers are seen, only the difference is pushed and pulled, making your deploys faster and saving you disk space. With a save/load command, all layers are sent every time since you do not know the state of the remote server when you run the save.
Second, this doesn't scale as you add hosts to run images. Every host would need the image copied on the chance you want to run it on that host, e.g. to handle failover or load balancing. It also won't work if you move to swarm mode or kubernetes since you could easily add new nodes to the cluster that won't have your image. Swarm mode defaults to looking up the sha256 of the image on the registry to guarantee the same image is always used even if the tag is modified on the registry after the initial deploy.
Keep in mind you can run your own registry server (there's a docker image and the api is open). Many artifact repositories (e.g. artifactory and nexus) include support for a docker registry. And many cloud providers include a registry with their container offerings. So you do not need to push to a remote docker hub to deploy locally.
One last point is that a registry server is useful to developers who can now pull the same image used in dev and prod to test against other microservices they are writing locally without the need to build everything locally or ssh to a CI/CD server or even prod to save and scp images back to their laptops.
Usually, you use a CI, CD pipeline when you want to streamline your build / test/ deploy process, and usually this happens if you have a production infrastructure to maintain that is actually critical to your business.
There is no need for a CI/CD pipeline if you're just playing around / prototyping IMO, in which case you can build you docker images on the machine directly, or ssh an image over. That's perfectly reasonable.
Look at the 'registry' as a repository of your binary image (i.e. a fixed version of your code that ideally is versioned and you know works)
Then deploying is as simple as telling your servers to pull the image and run it, from anywhere.
On a flexible architecture, you might have nodes coming up or going down at any time, and they need to be able to pull the latest code from somewhere to get back up and running automatically, at any time, without intervention.
Registry is single source of truth in this case. It means, that you can have multiple nodes (servers), cluster(s) and have the single place from where you can get your images. Also if of your nodes drop-down - you can fast start your image in the new one. Also you can automate your image's updating using registry's webhook, for example when you add new version of image registry gonna send webhook to any service that can upgrade your containers to the newest version.
Consider docker image as a new way of distribution of your software to your servers and docker-registry as a centralized storage of shared images(the like npm.org for js, maven.org for java).
For example,
if you develop java application, years before docker you may use .jar files to do it. The way docker image is better is that also include all OS level dependencies like JDK/JRE and system configurations. So this helps you to avoid "it works on my machine" effect.
To distribute docker image you can also use just docker file and build it all the time on every machine. Docker-Repository allows you to have centralized storage of pre-build images.
Pushing to docker-repository in your CI/CD allows to build your distributive once and further work with the same distributive both on integration and prod environments.
Using just Dockerfile will not guarantee you the same state on every build in every moment of time because you may install external dependencies in your Dockerfile script which may be updated or even removed between two sequential builds.
Although most of the time I am developing Java apps and am simply using Maven so my builds should be reproducible (at least that's what Maven says).
But say you are compiling a C++ program or something a little more involved, should you build inside of docker?
Or ideally use vagrant or another technology to produce reproduce able builds.
How do you manage reproducible build with docker?
You can, but not in your final image, as that would mean a much larger image than necessary: it would include all the compilation tool, instead of limiting to only what you need to execute the resulting binary.
You can see an alternative in "How do I build a Docker image for a Ruby project without build tools?"
I use an image to build,
I commit the resulting stopped container as a new image (with a volume including the resulting binary)
I use an execution image (one which only contain what you need to run), and copy the binary from the other image. I commit again the resulting container.
The final image includes the compiled binary and the execution environment.
I wanted to post an answer to this as well actually because to build on VonC's answer. Actually I just had Redhat Openshift training and they use a tool called Source to Image s2i, which uses docker to create docker images. And actually this strategy is great for managing a private (or public) cloud, where your build may be compiled on different machines, but you need to keep the build environment consistent.
(I understand this question is somewhat out of scope for stack overflow, because contains more problems and somewhat vague. Suggestions to ask it in the proper ways are welcome.)
I have some open source projects depending in each other.
The code resides in github, the builds happen in shippable, using docker images which in turn are built on docker hub.
I have set up an artifact repo and a debian repository where shippable builds put the packages, and docker builds use them.
The build chain looks like this in terms of deliverables:
pre-zenta docker image
zenta docker image (two steps of docker build because it would time out otherwise)
zenta debian package
zenta-tools docker image
zenta-tools debian package
xslt docker image
adadocs artifacts
Currently I am triggering the builds by pushing to github and sometimes rerunning failed builds on shippable after the docker build ran.
I am looking for solutions for the following problems:
Where to put Dockerfiles? Now they are in the repo of the package needing the resulting docker image for build. This way all information to build the package are in one place, but sometimes I have to trigger an extra build to have the package actually built.
How to trigger build automatically?
..., in a way supporting git-flow? For example if I change the code in zenta develop branch, I want to make sure that zenta-tools will build and test with the development version of it, before merging with master.
Are there a tool with which I can overview the health of the whole build chain?
Since your question is related to Shippable, I've created a support issue for you here - https://github.com/Shippable/support/issues/2662. If you are interested in discussing the best way to handle your scenario, you can also send me an email at support#shippable.com You can set up your entire flow, including building the docker images, using Shippable.
I am gaining knowledge about Docker and I have the following questions
Where are Dockerfile's kept in a project?
Are they kept together with the source?
Are they kept outside of the source? Do you have an own Git repository just for the Dockerfile?
If the CI server should create a new image for each build and run that on the test server, do you keep the previous image? I mean, do you tag the previous image or do you remove the previous image before creating the new one?
I am a Java EE developer so I use Maven, Jenkins etc if that matter.
The only restriction on where a Dockerfile is kept is that any files you ADD to your image must be beneath the Dockerfile in the file system. I normally see them at the top level of projects, though I have a repo that combines a bunch of small images where I have something like
top/
project1/
Dockerfile
project1_files
project2/
Dockerfile
project2_files
The Jenkins docker plugin can point to an arbitrary directory with a Dockerfile, so that's easy. As for CI, the most common strategy I've seen is to tag each image built with CI as 'latest'. This is the default if you don't add a tag to a build. Then releases get their own tags. Thus, if you just run an image with no arguments you get the last image built by CI, but if you want a particular release it's easy to say so.
I'd recommend keeping the Dockerfile with the source as you would a makefile.
The build context issue means most Dockerfiles are kept at or near the top-level of the project. You can get around this by using scripts or build tooling to copy Dockerfiles or source folders about, but it gets a bit painful.
I'm unaware of best practice with regard to tags and CI. Tagging with the git hash or similar might be a good solution. You will want to keep at least one generation of old images in case you need to rollback.