I can't find docs anywhere describing whether the "FROM" Dockerfile command will automatically pull from a remote private registry, or whether you must do an explicit "docker pull" first before running "docker build".
I have an insecure private registry set up at http://192.168.1.1:5000. It is configured in /etc/docker/daemon.json. Pushes and pulls both work against this insecure registry.
But if an image doesn't exist in the local cache (i.e., it hasn't been pulled and doesn't already exist from a previous docker build) then when running docker build the Dockerfile FROM command results in:
=> ERROR [internal] load metadata for 192.168.1.1:5000/reponame 0.1s
------
> [internal] load metadata for 192.168.1.1:5000/reponame:latest:
------
failed to solve with frontend dockerfile.v0: failed to create LLB definition: failed to do request: Head "https://192.168.1.1:5000/v2/reponame/manifests/latest": http: server gave HTTP response to HTTPS client
Strange that it is trying HTTPS even though it should know that the registry is insecure and thus use HTTP.
Anyway, is this normal behavior? Do I just need to always perform a "docker pull" for an image before attempting to reference it in a Dockerfile FROM command? Or am I missing something that would make this work without doing a pull first?
Posting answer for anyone who might run into this in the future. I finally found the root cause of this issue. Apparently the standard docker registry image doesn't respond to the http HEAD requests that DOCKER_BUILDKIT sends.
The workaround is to disable buildkit (set DOCKER_BUILDKIT=0 in environment). This falls back to the pre-buildkit logic, which doesn't make that HEAD call before pulling the image. Less efficient (and you lose some other buildkit advantages), but it works.
Alternatively, you can issue a pull command for the image before running the docker build, and FROM will pull from the local cache.
Related
Try to login to login to my private compagny docker (jfrog) registry, it was working ...
[rabxxxx.xxx#vmxxx]$ docker login registry.x.xx-xx-xxxxx.xx.xx.xxxx.net -u xxx
Password:
WARNING! Your password will be stored unencrypted in /home/rabxxxx.xxx/.docker/config.json.
Configure a credential helper to remove this warning. See
https://docs.docker.com/engine/reference/commandline/login/#credentials-store
Login Succeeded
[rabxxxx.xxx#vmxxx]$ docker run -p 8080:8080 -p 50000:50000 jenkins
Unable to find image 'jenkins:latest' locally
docker: Error response from daemon: Get "https://registry-1.docker.io/v2/": context deadline exceeded.
See 'docker run --help'.
cat /etc/docker/daemon.json
{
"insecure-registries": ["registry.x.xx-xx-xxxxx.xx.xx.xxxx.net"],
"registry-mirrors": ["registry.x.xx-xx-xxxxx.xx.xx.xxxx.net"]
}
Please can you tell what I must check and why it's return me this
Error response from daemon: Get "https://registry-1.docker.io/v2/"
Thx
docker pull registry.x.xx-xx-xxxxx.xx.xx.xxxx.net/jenkins`
Using default tag: latest Error response from daemon: manifest for registry.repo.proxy-dev-forge.asip.hst.fluxus.net/jenkins:latest not found: manifest unknown: The named manifest is not known to the registry.
( same with /library/jenkins ]
When you docker run jenkins, docker will expand that to docker.io/library/jenkins, where docker.io is Docker Hub and the actual underlying registry server is registry-1.docker.io. When you define a mirror for that, you need to keep the repository names the same, so not registry.example.org/jenkins, but registry.example.org/library/jenkins.
In your case, it appears your mirror has neither of those, making it an incomplete mirror, assuming it's a mirror at all. But you don't really want to run the library/jenkins image anyway, since they've indicated it's deprecated with a recommendation to use the jenkins/jenkins repo instead (likely maintained by Cloudbees).
Please can you tell what I must check and why it's return me this
Error response from daemon: Get "https://registry-1.docker.io/v2/"
Mirroring in docker is a best effort. If the mirror fails for any reason, or you try to do a push instead of a pull, docker will fall back to going direct to Docker Hub.
If you only want to talk to your local registry, then you should put that registry in all your image names. This is particularly important if your mirror doesn't match Docker Hub and you make assumptions based on your local registry. In those cases, the same definition for an image on one machine may run very different code if the network has an issue or you deploy on a different machine without the mirroring specification. RedHat documented many of these risks after pushing hard to add the ability to overload the top level namespace, and we had a wave of dependency confusion attacks impact other software repositories.
Your company registry did not cache image from official registry. Try to make it auto cache or pull the jenkins image from docker official registry and push to your company registry then you can use it. I will suggest you to make the company registry auto cache, when the registry exists in your private registry, just use it, if did not exists, cache from the official registry.
I am trying to pull docker image from Nexus repo without using the registry mirror in the command line and it is throwing an error. If I use the registry mirror in the pull it is succeeding but the image name is not I would like.
My docker version is:
Docker version 20.10.8, build 3967b7d
My nexus version is
Sonatype Nexus Repository ManagerOSS 3.31.1-01
docker system info:
Insecure Registries:
xxx.xxx.x.xxx:8083
127.0.0.0/8
Registry Mirrors:
http://xxx.xxx.x.xxx:8083/
When I run: sudo docker pull xxx.xxx.x.xxx:8083/mongo:4.2.3, it succeeds and the debug info is:
DEBU[2021-08-17T10:37:19.364681226-04:00] Calling HEAD /_ping
DEBU[2021-08-17T10:37:19.365301100-04:00] Calling POST /v1.41/images/create?fromImage=192.168.9.175%3A8083%2Fmongo&tag=4.2.3
DEBU[2021-08-17T10:37:19.367151579-04:00] Trying to pull xxx.xxx.x.xxx:8083/mongo from https://xxx.xxx.x.xxx:8083 v2
WARN[2021-08-17T10:37:19.374915464-04:00] Error getting v2 registry: Get https://xxx.xxx.x.xxx:8083/v2/: http: server gave HTTP response to HTTPS client
INFO[2021-08-17T10:37:19.374944418-04:00] Attempting next endpoint for pull after error: Get https://xxx.xxx.x.xxx:8083/v2/: http: server gave HTTP response to HTTPS client
DEBU[2021-08-17T10:37:19.374964188-04:00] Trying to pull xxx.xxx.x.xxx:8083/mongo from http://xxx.xxx.x.xxx:8083 v2
DEBU[2021-08-17T10:37:19.398630498-04:00] Fetching manifest from remote digest="sha256:92814bb60dc673bb68b6aca0b24bcb8738d7b2c267b97ce62fa92adc3746a0ea" error="<nil>" remote="192.168.9.175:8083/mongo:4.2.3"
DEBU[2021-08-17T10:37:19.429454057-04:00] Pulling ref from V2 registry: xxx.xxx.x.xxx:8083/mongo:4.2.3
When I run: sudo docker pull mongo:4.2.3 it fails to pull the image from Nexus with an error and pulls from docker.io on the next try. Debug info as below:
DEBU[2021-08-17T10:26:25.078886904-04:00] Calling HEAD /_ping
DEBU[2021-08-17T10:26:25.079306196-04:00] Calling GET /v1.41/info
DEBU[2021-08-17T10:26:25.097994642-04:00] Calling POST /v1.41/images/create?fromImage=mongo&tag=4.2.3
DEBU[2021-08-17T10:26:25.099642151-04:00] Trying to pull mongo from http://xxx.xxx.x.xxx:8083/ v2
INFO[2021-08-17T10:26:25.116000813-04:00] **Attempting next endpoint for pull after error: manifest unknown: manifest unknown**
DEBU[2021-08-17T10:26:25.116039299-04:00] Trying to pull mongo from https://registry-1.docker.io v2
DEBU[2021-08-17T10:26:25.305043063-04:00] Fetching manifest from remote digest="sha256:58b25d51baa11a85b6aedf7c4e05710d12a27ddc2883e2692e7d58527d98bd73" error="<nil>" remote="docker.io/library/mongo:4.2.3"
DEBU[2021-08-17T10:26:25.360955030-04:00] Pulling ref from V2 registry: mongo:4.2.3
DEBU[2021-08-17T10:26:25.361036645-04:00] docker.io/library/mongo:4.2.3 resolved to a manifestList object with 5 entries; looking for a unknown/amd64 match
Issue with Image name:
REPOSITORY TAG IMAGE ID CREATED SIZE
xxx.xxx.x.xxx:8083/mongo 4.2.3 97a9a3e85158 17 months ago 386MB
Any guidance on this would help.
Nexus Docker ( xxx.xxx.x.xxx:8083) is pointed to hosted Type on port 8083 and the mongo:4.2.3 is uploaded into this docker type. We ultimately want to use this in a air gapped system where there is no internet connection.
There are three things going on here:
I am trying to pull docker image from Nexus repo without using the registry mirror in the command line and it is throwing an error. If I use the registry mirror in the pull it is succeeding but the image name is not I would like.
I'm going to recommend changing your likes. :)
If you want to pull from a specific registry, then use that registry in the image name. Trying to refer to your local registry with short names is merging two different image registry namespaces, which means it's trivial to run an image from the wrong namespace and result in a security breach. This was a large issue for other package repositories (see "dependency confusion" attacks) that docker was not susceptible to because they require the registry name as part of the image name (the only exception being Docker Hub). Even RedHat who tried to get options like add-registry and block-registry into the upstream docker engine (and failed, these options only ever appeared in a RedHat specific fork) is now telling users that it was a very bad idea and now their users are exposed to security vulnerabilities they can't easily fix because removing the feature will break lots of user environments.
Next, why doesn't the pull go to your registry? Because your image name doesn't match that of Docker Hub. Official images without a username are actually under the library repository. This is typically hidden from view, but you can do things like docker pull library/alpine or even docker pull docker.io/library/alpine instead of docker pull alpine, and all 3 will be pulling from the same place.
The fix is to run
docker pull xxx.xxx.x.xxx:8083/mongo:4.2.3
docker tag xxx.xxx.x.xxx:8083/mongo:4.2.3 xxx.xxx.x.xxx:8083/library/mongo:4.2.3
docker push xxx.xxx.x.xxx:8083/library/mongo:4.2.3
The last issue I actually can't help you with, it comes from the error message you're seeing when pulling from Hub, which should work:
docker.io/library/mongo:4.2.3 resolved to a manifestList object with 5 entries; looking for a unknown/amd64 match
The unknown/amd64 is unexpected to me, typically that would be linux/amd64 so there is something unexpected with the platform you're running your commands on. If you want to get into debugging that, update your question with docker info. You can try working around that with:
docker pull --platform linux/amd64 mongo:4.2.3
to force the platform, but that still doesn't explain why it doesn't know your current platform.
I guess you are trying to set your nexus docker repository to be the default one for the machine in the sealed network.
that needs changing because of the following from docker documentation:
Tag an image for a private repository
To push an image to a private registry and not the central Docker registry you must tag it with the registry hostname and port (if needed).
$ docker tag 0e5574283393 myregistryhost:5000/fedora/httpd:version1.0
with more upfront configuration and upkeep but no changes requiered for the client machines
Is if you have a DNS server in your network you could point docker.io to your nexus host ip address and put a proxy to intercept the communication and redirect and adapt the requests as they were to the nexus docker registry
Hopes this solves your pickle :)
Update 1:
It could be that you need to also change /etc/containers/registries.conf like specified here to only or also specify your nexus docker registry.
Update 2:
Before letting Gopi give up entirely, I would suggest using Podman as an alternative to Docker. Podman is a daemon-less container engine that works by forking processes to handle each running container. It seamlessly works with docker images thanks to the OCI standard, and on top of that, the only change when using it is replacing the docker command prefix with podman since all the commands are exactly the same. Podman was created by RedHat so by default it searches RedHat repos and you can add your own too as shown in this article that I mentioned before.
I'm trying to set a docker mirror to be the default mirror to pull/push images.
As per documentation I already set the file /etc/docker/daemon.json with the following:
{
"registry-mirrors": ["https://localregistry"]
}
Then I try the following:
docker login localregistry
docker pull localregistry/image:tag > it works
docker pull image:tag > doesn't work
I'm always getting "no basic auth credentials error" from the docker daemon, but from the registry log I get err.code="manifest unknown" err.detail="unknown tag"
Any idea?
I'm using docker version 19.03.08
docker login localregistry
First, I hope this is changing the name for the question, because the registry name localregistry will not work...
docker pull localregistry/image:tag > it works
The fact that this works indicates that you likely have a registry name with a . or : in the hostname. Otherwise docker would try to pull localregistry/image:tag from the localregistry user on Docker Hub.
docker pull image:tag > doesn't work
This should always work, failures should be transparent to the user if it's really a mirror of Docker Hub. What happens is it resolves that name to docker.io/library/image:tag, first tries to pull from localregistry/library/image:tag, and any error falls back to a pull from Docker Hub, and any error there finally shows to the user.
Most likely the issue is that you didn't include library as the repo name for your image in the local registry.
If you are using this to include images that don't exist on Docker Hub, then I would skip the mirror and simply refer to the mirror explicitly. Doing otherwise creates many opportunities for nonintuitive failures that aren't easy to see. E.g. a stale image can be pushed to your mirror in place of an upstream image, and Docker will stop pulling updates from upstream. And because any mirror errors fall back to Hub, if you use an image name that you have no control over upstream, someone else could take that name on Hub and begin injecting unknown or even malicious images into your server.
If this doesn't answer your question, then I'd recommend using your question with actual image names and error messages from the logs showing what specifically failed (you can mask out part of the registry name of necessary).
I'm in low cost project that we send to container registry (DigitalOcean) only latest image.
But all time, after running:
docker build .
Is generating the same digest, every time.
This is my script for build:
docker build .
docker tag {image}:latest registry.digitalocean.com/{company}/{image}:latest;
docker push registry.digitalocean.com/{company}/{image}
I tried:
BUILD_VERSION=`date '+%s'`;
docker build -t {image}:"$NOW" -t {image}:latest .
docker tag {image}:latest registry.digitalocean.com/{company}/{image}:latest;
docker push registry.digitalocean.com/{company}/{image}
but not worked.
Editing my answer, what David said is correct - the push with out the tag should pick up latest tag.
If you provide what you have in your local repository and the output of the above commands, it would shed more light to your problem.
Edit 2:
I think I have figured out on why:
Is generating the same digest, every time.
This means, although you are running your docker build - there has been no change to the underlying artifacts which are being packaged into the image and hence it results into the same digest.
Sometimes layers are cached but there are changes that aren't detected so you can delete the image or use 'docker system prune' to force clearing cache here
I'm unable to pull docker images in my environment. I think it's blocked by company firewall, but I'm not sure why It gets layer info and later It prints that repository is not found.
sudo docker pull hello-world
latest: Pulling from hello-world
50a54e1f9180: Pulling fs layer
7a5a2d73abce: Pulling fs layer
Pulling repository hello-world
Repository not found
Docker version: (I cannot upgrade to newest docker on RHEL 6.9)
Docker version 1.7.1, build 786b29d/1.7.1
Could somebody explain me which protocols (https only?) are used during docker image pulling phase and what addresses are contacted ("https://registry-1.docker.io/v2" only?) ?
Docker images can consist of multiple layers. By default, the Docker daemon will pull three layers of an image at a time but will pull less in case an image has lesser layers. Also, if no tag is provided, Docker Engine uses the :latest tag as a default. Above is a basic log of your pull request indicating docker trying to pull layers of the image but failing, may be due to firewall restrictions or older docker version.
Docker uses the https:// protocol to communicate with a registry,
unless the registry is allowed to be accessed over an insecure
connection.
Not sure what all addresses it tries to connect to pull an image.
Problem was that firewall was blocking connections during pulling images.
Docker registry uses CDN so more URLs need to be allowed and not only registry URL.
I requested to allow the following URLs on company firewall and it is working now.
dseasb33srnrn.cloudfront.net
auth.docker.io
elb-registry.us-east-1.aws.dckr.io.
us-east-1-elbregis-10fucsvj1tcgy-133821800.us-east-1.elb.amazonaws.com
registry-1.docker.io
registry-origin.docker.io
index.docker.io
elb-io.us-east-1.aws.dckr.io
us-east-1-elbio-rm5bon1qaeo4-623296237.us-east-1.elb.amazonaws.com
Docker log file (/var/log/docker) help me to identify root problem.
There were the following errors:
level=error msg="Error from V2 registry: Get https://dseasb33srnrn.cloudfront.net/registry-v2/docker/registry/v2/blobs/sha256/78/78445dd45222097f5f8d5a16e48dc19c4ca162dcdb80010ab6f1ccfc7e2c0fa3/data?Expires=1493033299&Signature=DiEmffSxF1F9z-SRoGyX3NwzfeQY3BhE2Du3aPb1qy9VglXyn1mus7Xy9Y~DQnwaQ9IIN71FboK5lOAiN1Qj-x662qhioi72CJ-v02fiMHqC03FDb0l4LyULquU8GaalW3uZG4hdfuSqOBQ1qo9HEcxhMyQGqOqpfPUKjUlHqm8_&Key-Pair-Id=APKAJECH5M7VWIS5YZ6Q: read tcp 52.85.173.110:443: connection reset by peer"
The list of URLs which needs to be allowed I found here:
https://forums.docker.com/t/list-of-docker-hub-mirror-sites-to-configure-proxy-whitelist/20845/2