Docker build seems to be stable on arm64 now [1], after implementing the above caching mechanism proposed + moving the docker build jobs from `aarch64_dev` to `aarch64_build` jenkins slaves (to eliminate the overlapping of IEC fuel/compass deploy/testing jobs with the validation docker build job, which used to cause issues due to deploy jobs restarting the Docker service on the host).

[1] https://jenkins.akraino.org/view/validation/job/validation-docker-build-arm64-master/

The error in this ticket does not seem to be caused by an incomplete PATH, but rather out-of-memory spurious failures during database fetch/update using `go-cve-dictionary fetchnvd`.

To bypass the issue (considering a ~50% reproduction rate of the issue during manual testing with a ~1h test iteration duration, debugging and rootcausing this is non-trivial), we chose to leverage the fact that all affected builds are scheduled on static Jenkins slaves, so we can cache the databases locally on the build server between container builds, reducing the fetch size considerably and thus avoiding OOM issues.

I proposed a patch for implementing this persistent cache and will manually initialize the DB cache on the `aarch64_dev` jenkins slaves.

Hi Juha, you are correct, this should not be closed yet. The build passed on its own a couple of days after the Jira was opened, and I thought it's fixed. But looking at the jobs, this seems to be an intermittent error. It might have to do something with the servers we build the images on, not with the code itself. We'll continue to investigate what's going on.

cristinapauna The patch referred has not been merged; is this corrected somewhere else or correction not needed at all?

https://gerrit.akraino.org/r/c/validation/+/2294