Container deploy not deploying the latest version of an image tag

Questions and Answers Deployment
1

Hello, I’m having an issue where Qovery won’t deploy the latest version of my image when the tag has already been deployed previously (eg. backend:production).

Workflow :

  • from Github Action CI,
  • when a production deployment workflow is triggered
  • I tag the to-be-deployed images with the production tag on my image repository (ie. AWS ECR)
  • using the Qovery CLI, I instruct Qovery to deploy my service using the tag production

During the previous deploy Qovery has already deployed my service with the production tag, but during a new deploy, the image tagged with production is a new and different image.
However Qovery seems to be deploying a previous image.

Is my backend:production image cached at the Qovery level (ie. Qovery image mirroring process qovery-mirror-2274d177-cf13-4316-ac22-bfc52abf9312)?

What can I do to ensure that Qovery deploy the latest image using my production tag?

More info :

My container service setting on Qovery
Settings > General > Container settings > Image Tag
seems to be updated each time I deploy using the CLI

            qovery container deploy \
              --organization my-org \
              --project my-org-monorepo \
              --environment prod \
              --container "backend" \
              --tag production \
              --watch

In the meantime I updated the workflow to use the git commit hash of my images to prevent Qovery deploying an older version of my image (eg. backend:2274d1c)

Thank you!
cc @Morgan_Perry

1 Like
2

Hi Benjamin,

as you have guessed, there is an image mirroring on the image registry created by Qovery within your cloud account to ensure we always have the image available (for deployment but also if Kube needs it). You can find more info on the pipeline in this article.

image
image1678×875 120 KB

In general, we do not encourage to use of tags like latest or production and instead use unique image tags so that there will be no caching issues along the pipeline.

Alessandro

1 Like