Problem:

Server version 4.8.x had an issue preventing the correct version of machine-agent from being pinned causing a mismatched machine-agent version to be used.

Incompatibility of the mismatched machine-agent blocks successful job startup.

Solutions:

• Solution 1 (with a bulleted list, if applicable)

  • kubectl edit deployment [-n namespace] machine-provisioner-provisioner
    • Delete the environment variable ENTERPRISE_USE_PAIRED_AGENT_VERSION
    • Add the variable ENTERPRISE_PINNED_AGENT_VERSION with value "1.0.101561-345563e1"

• Solution 2

  • Deploy the latest patch release 4.8.6 which will correct the version pinning issue.

Outcome:

After editing the deployment, machine-provisioner-provisioner pods will be rolled with the correction and your machine executors should run as normal.

Additional Notes:

Manually modifying the deployment environment variables is a temporary workaround that will be undone by a helm deploy. If a helm upgrade is performed the workaround will need to be re-applied.

Once a corrected point version of 4.8.x has been released with the fix and deployed the temporary workaround will no longer be required.

Additional Resources:

• CircleCI Server Changelog