Google Distributed Cloud Virtual Storage simplified storage

Storage simplified by managing policies, Google Distributed Cloud Virtual Storage

Customers can run the same Kubernetes as Google Kubernetes Engine in the cloud on their own hardware and data centers using GKE Enterprise capabilities to manage clusters at enterprise scale with GDCV vSphere. The scalable, secure, and highly available architecture of GDCV (vSphere) supports enterprise business applications.

Google Distributed Cloud Virtual Storage simplifies on-premises Kubernetes deployment by connecting with VMware vSphere. The GDCV (vSphere) adds HA admin and user clusters, auto-scaling, node repair, and VMware’s sophisticated storage infrastructure to VMware vSphere for high availability.

Problem

Virtual disk deployment is automated by many VMware customers using datastore clusters. By merging many datastores into one object, they may let vSphere choose the optimum virtual disk location for the particular requirement.

Container Storage Interface (CSI) driver module controls Kubernetes-vSphere storage interactions. VSphere CSI is VMware’s CSI driver. Since the VMware-delivered driver does not support datastore clusters, this integration is fast but limited.

VMware uses Storage Policy Based Management (SPBM) to declare workload datastore clusters and offer placement logic for automatic storage placement. Since SBPM was not supported in GDCV, VM managers used to SBPM for VMs found storage on these clusters tougher and less obvious.

Solution

GDCV (vSphere) 1.16 supports SPBM, allowing customers to declare datastore clusters and deploy workloads consistently. SPBM lets GDCV manage vSphere storage without affecting Google Distributed Cloud Virtual Storage or Kubernetes. GDCV lifecycle management relies on contemporary storage integration on vSphere for greater resilience and a shorter planned maintenance window.

GDCV interaction with VMware’s SPBM and improved VMware CSI driver use enable this new storage assignment architecture. SPBM provides a single control plane for a variety of data services and storage systems. The framework aligns virtual machine storage with application needs. Create storage rules mapping VMs or applications to their storage demands with SPBM.

By integrating with VMware’s SBPM, building Google Distributed Cloud Virtual Storage clusters with storage is now as simple as referencing a storage policy in the install configurations.

Like datastore clusters, storage policies can handle several datastores as one. The policy returns a list of all conforming datastores but does not recommend one for placement. For that, GDCV will handle it. It analyzes all compatible datastores in a policy and chooses the best disk placement datastore. For each storage placement, SPBM with GDCV does that dynamically.

When storage capacity or maintenance changes, all adjustments are made from the storage end, which is great for automation. Adding storage capacity no longer requires modifying Google Distributed Cloud Virtual Storage configuration files.

This simplifies Google Distributed Cloud Virtual Storage storage management a much.

Reviewing SBPM policies

A VMware administrator can create storage policies depending on the storage array’s capabilities with SBPM. They can map one/many datastores to one/many policies and assign the VM the optimum storage policy. In practice,They construct a policy specifying all gold level storage (e.g. SSD exclusively for production) and add all matching datastores to it. This policy is then assigned to VMs. Bronze-level storage too. Simply deploy a Bronze storage policy to the required VMs with bronze datastores (e.g. HDD exclusively for Dev environment).

In GDCV, SPBM works

To use the storage policy feature in Google Distributed Cloud Virtual Storage, the VMware admin must put up at least one storage policy compatible with one or more datastores your cluster may access.

You can place datastore tags with GDCV. VMware managers can tag storage depending on performance levels, cluster name, and disk kinds.

Let’s see how SBPM policies can govern VM disk allocation and persistent volume claims (PVC) for stateful applications in a cluster.

VM disk location

Consider her Gold and Bronze samples again. Storage needs for Google Distributed Cloud Virtual Storage User clusters will be defined. The User cluster configuration file has two storage policies:

• The cluster’s default storage policy, our “Bronze Policy,”
• Storage policy for user cluster node pools our “Gold” strategy
• The GDCV node VMs’ datastores receive the “Gold” and “Bronze” tags first. In this scenario, “Gold” refers only to SSDs and “Bronze” to HDDs.
• Create and assign tags using the documentation. Tags can also be applied to datastore clusters or datastores inside them.

Following the official guidelines, the storage policy is defined once tags are generated.

A user cluster config file excerpt sets the storage policy name “Bronze” at the cluster level. All provisioned VMs in all nodepools will utilize this storage policy to discover compatible datastores and dynamically select the one with enough capacity.

A node-pool storage policy (“Gold”) is in this user cluster config file excerpt. VMs will be provisioned at that node pool using this policy, while all other storage will follow the cluster section policy.

This storage policy hides storage details from cluster admin. If capacity is an issue, the VMware admin can tag extra datastores to add to the storage policy. The cluster automatically integrates the policy-provided capacity without any action from the Google Distributed Cloud Virtual Storage cluster admin. This reduces cluster admin workload and automates storage management.

Continued Volume Claims

A default StorageClass can be one of several StorageClass objects in a user cluster. When you create the cluster using the install guide, They establish a default storage class.

Default storage classes can be replaced by others. Google Distributed Cloud Virtual Storage user clusters can utilize the vSphere CSI driver to create storage classes that reference any vCenter storage policy.

This means that PVC-created volumes will be dispersed among datastores that support our user clusters’ storage policies. These classes map to vSphere VMFS, NFS, and vSAN storage rules.

Conclusion

This post discusses Google Distributed Cloud Virtual Storage integration with VMware SPBM. This integration automates storage management by moving it away from hard linkages between datastores and toward VMware-managed dynamic storage assignment. More storage management flexibility and minimal overhead and downtime for GDCV clusters.