How to Prevent Container Runtime From Being Managed by Platform9
Problem
In a Platform9 Managed Kubernetes (PMK) deployment, each node is running an instance of the pf9-__nodeletd service which is used to bootstrap the cluster, perform health checks, and manage its overall lifecycle – including underlying components like the container runtime, e.g. Docker.
If a critical failure is incurred – some components, including the container runtime, may need to be re-initialized. In this case, any pods scheduled to the node would be affected (as per the underlying containers having been destroyed as part of the container runtime being initialized).
In some cases, it may be preferred that the container runtime be managed outside the control of PMK – including installation of the binary.
Environment
- Platform9 Managed Kubernetes - All Versions
- Container Runtime (Docker)
Procedure
Docker
- In / etc/pf9/kube_override.env, dictate that the PF9_MANAGED_DOCKER variable be exported as " false".
Note: The override file will be sourced on the next restart of the pf9-__nodeletd service, which will automatically occur (based on failures as aforementioned). As such, there is no need to manually restart the service after making the change.