5 ms·

I've tried one of the preview builds as part of their Developer Preview Program[1] and it's super fast when you're running an arm64 image, as opposed to an amd6

I've tried one of the preview builds as part of their Developer Preview Program[1] and it's super fast when you're running an arm64 image, as opposed to an amd64 that will automatically be run via qemu (and therefore incur some overhead). By "super fast", I mean I was seeing speeds comparable to running the commands natively on the M1 mac (npm install and a heavy Gulp build). [1] - https://www.docker.com/community/get-involved/developer-preview https://www.docker.com/community/get-involved/developer-prev...

Axsuul 6y ago
How close are we to be able to deploy arm64 images to production/AWS/GCP?
- justincormack 6y ago
  You can deploy to AWS Graviton machines.
- Jonnax 6y ago
  You can have done this for about 2 years.
- dividedbyzero 6y ago
  Seems like GCP isn't offering this currently. It's available on AWS EC2: https://aws.amazon.com/ec2/graviton/ https://aws.amazon.com/ec2/graviton/
vocram 6y ago
Are users able to check if they are running a native arm64 container image, as opposed to an amd64 emulated one? Or that’s something you have to infer from the execution speed?
- gmaster1440 6y ago
  You can use manifest inspect[1] to see all supported architectures, and you should also be able to see on Docker Hub[2] as well. EDIT: If you're asking specifically if you can tell what architecture a running container is using, a simple `docker inspect CONTAINER_ID` should show it. [1] https://docs.docker.com/engine/reference/commandline/manifest_inspect/ https://docs.docker.com/engine/reference/commandline/manifes... [2] https://hub.docker.com/_/node https://hub.docker.com/_/node
  - vocram 6y ago
    I just checked myself, and you get notified with a nice WARNING upon a `docker run`: WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested
- LukeLambert 6y ago
  I'm not great at bash, but this little script worked for me. #!/usr/bin/env bash docker ps --format "{{.ID}} {{.Names}}" | while read line; do read -ra arr <<<"$line" id="${arr[0]}" names="${arr[1]}" arch="$(docker exec "$id" uname -m)" echo -e "$arch\t$names" done Or this one-liner with worse formatting: docker ps --format "{{.Names}}" | xargs -tI {} docker exec {} uname -m

Axsuul 6y ago

How close are we to be able to deploy arm64 images to production/AWS/GCP?

justincormack 6y ago
You can deploy to AWS Graviton machines.
Jonnax 6y ago
You can have done this for about 2 years.
dividedbyzero 6y ago
Seems like GCP isn't offering this currently. It's available on AWS EC2: https://aws.amazon.com/ec2/graviton/ https://aws.amazon.com/ec2/graviton/

vocram 6y ago

Are users able to check if they are running a native arm64 container image, as opposed to an amd64 emulated one? Or that’s something you have to infer from the execution speed?

gmaster1440 6y ago
You can use manifest inspect[1] to see all supported architectures, and you should also be able to see on Docker Hub[2] as well. EDIT: If you're asking specifically if you can tell what architecture a running container is using, a simple `docker inspect CONTAINER_ID` should show it. [1] https://docs.docker.com/engine/reference/commandline/manifest_inspect/ https://docs.docker.com/engine/reference/commandline/manifes... [2] https://hub.docker.com/_/node https://hub.docker.com/_/node
- vocram 6y ago
  I just checked myself, and you get notified with a nice WARNING upon a `docker run`: WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested
LukeLambert 6y ago
I'm not great at bash, but this little script worked for me. #!/usr/bin/env bash docker ps --format "{{.ID}} {{.Names}}" | while read line; do read -ra arr <<<"$line" id="${arr[0]}" names="${arr[1]}" arch="$(docker exec "$id" uname -m)" echo -e "$arch\t$names" done Or this one-liner with worse formatting: docker ps --format "{{.Names}}" | xargs -tI {} docker exec {} uname -m