Should I Use the RWO or RWX Access Mode for Scaling Workloads in GKE?

Question

I want to scale workloads in Google Kubernetes Engine (GKE) using Kubernetes autoscaling. Should I configure Persistent Volume Claims (PVCs) to use the ReadWriteOnce (RWO) or ReadWriteMany (RWX) access mode?

Environment

• HVR 6
• Google Kubernetes Engine (GKE)

Answer

ReadWriteOnce

A PersistentVolume (PV) with the ReadWriteOnce access mode can be mounted by only one node at a time. This mode is suitable for single-node workloads that require data persistence.

If you use a ReadWriteOnce volume in a multi-node setup, your pods may get stuck in the ContainerCreating state with a Multi-Attach error. This happens because Kubernetes cannot mount the same ReadWriteOnce volume on multiple nodes.

ReadWriteMany

A PV with the ReadWriteMany access mode can be mounted by multiple nodes at the same time. This allows multiple pods to read from and write to the volume concurrently, which is ideal for scalable workloads, like HVR on GKE.

Not all GKE storage solutions in GKE support the ReadWriteMany access mode. Consider the following solutions:

• Google Cloud Filestore: A managed Network File System (NFS) that supports ReadWriteMany.
• NFS Server: Implement an NFS server that provides ReadWriteMany access.

Recommendation

Before switching to ReadWriteMany access:

1. Verify that your HVR configuration supports ReadWriteMany storage without affecting performance or data integrity.
2. Confirm compatibility with your application's workload and persistence architecture.
3. Test thoroughly in a staging environment before full deployment.