HCC Catalog (Helm Chart) vs. “Classic” Approach to Trident deployment

On NetApp HCI, use whatever the official documentation suggests (likely the Helm Chart (Catalog) approach).

SolidFire and eSDS users can use either approach.

• Catalog approach
  • Standardized for Rancher with both NetApp HCI and SolidFire
  • Good for frequent or repetitive deployments
  • Not all Trident versions may be available but versions recommended by NetApp HCI for use with Rancher should be available
  • Well-tested updates with documented update procedures specific to NetApp HCI
• Classic approach
  • You can choose to deploy any Trident version (which can be both good and bad; at the very least you should check which Trident version is supported by NetApp HCC - this information could be found in the Trident documentation or by checking the versions deployed by ez-rancher Trident Catalog). In theory at least any Trident version that supports Kubernetes contained in a Rancher version supported by NetApp HCC should be acceptable. But if you want to be on the safe side, check what versions are supported by NetApp HCC with Rancher and use one of those.
  • Somewhat manual process that can be automated, but also a source of problems if poorly planned and executed
  • Uses standard update procedures from NetApp Trident

NOTES:

• If you deploy multiple Rancher clusters against the same NetApp HCI or SolidFire back-end you must ensure that each Rancher cluster (and its Trident instance) use a different storage account so that CHAP authentication does not lead to one Rancher cluster seeing PVs from another
• When you destroy a cluster, remember to clean up the account, volumes, IQNs, VLANs, etc if you have no intent to use them again
• When assigning storage account names, assuming Rancher cluster have unique names, administrators could base SolidFire storage account names used by Trident upon Rancher cluster names

NetApp HCC Helm Chart for Workload Clusters

NetApp HCC users should be able to deploy it more easily.

ez-rancher users can deploy it or add it later. Find it here. Add this Helm Chart v3 to a User Cluster’s Catalog and deploy from the User Cluster’s Cluster Explorer (find it under Apps).

NetApp Trident “Classic”

This approach uses standard NetApp Trident deployment steps as documented here.

A YouTube video of the process with Rancher v2.5.2 and Trident v20.10 can be viewed here.

Customize Trident Deployment for Workload Clusters

• In Trident back-end configuration file, pick a unique (not trident) user name for the Trident account for each individual cluster
• On SolidFire array, use the same name or other appropriate convention to “link” cluster name with Trident name
• For example Rancher cluster rancher-dc1-az1-mfg-uc01 would use the same account name on SolidFire:

{
    "version": 1,
    "storageDriverName": "solidfire-san",
    "Endpoint": "https://admin:nimda@192.168.1.1/json-rpc/11.0",
...
    "TenantName": "rancher-dc1-az1-mfg-uc01",
}

Configure Trident QoS policies with non-overlapping Min-Max ranges

As long as this issue is open, it’s best to create Trident storage policies with non-overlaping Min-Max ranges unlike in this example:

• Slow: 200/2000/5000
• Fast: 500/5000/10000

This should work better:

• Slow: 200/1000/5000
• Fast: 2000/5000/10000