There appears to be a problem with the Helm deployment to Kind during the gaffer docker CI/CD pipeline. The exact issue is not clear but seems to be related to the accumulo master pod not starting correctly which has an obvious knock on effect to the rest of the deployment.
Some changes to the Helm chart have been made to make the data loader now dependant on a healthy accumulo master but what's causing this has not yet been bottomed out. Need some investigation to find and fix, however it may point to a wider issue with the helm charts. Debugging will require use of kubectl to inspect what is happening with the pods, it can be fairly frequently replicated by building the 2.2.0 images and running the k8s parts of the CI manually (e.g. deploy to kind script).
Note that the same accumulo image can be ran via docker just fine and produces a functional deployment using existing docker compose yaml.
Overview
There appears to be a problem with the Helm deployment to Kind during the gaffer docker CI/CD pipeline. The exact issue is not clear but seems to be related to the accumulo master pod not starting correctly which has an obvious knock on effect to the rest of the deployment.
Some changes to the Helm chart have been made to make the data loader now dependant on a healthy accumulo master but what's causing this has not yet been bottomed out. Need some investigation to find and fix, however it may point to a wider issue with the helm charts. Debugging will require use of
kubectlto inspect what is happening with the pods, it can be fairly frequently replicated by building the 2.2.0 images and running the k8s parts of the CI manually (e.g. deploy to kind script).Note that the same accumulo image can be ran via docker just fine and produces a functional deployment using existing docker compose yaml.