Description

This Knowledge Article provides comprehensive guidance on deploying Arctera InfoScale within OpenShift virtual machine (VM) environments, focusing on disk identification, network transport optimization, and fencing mechanisms to ensure reliable and high-performing cluster operations.

1.1 Serial Number Configuration Strategy

Proper disk identification is crucial for Veritas Volume Manager (VxVM) operations. Each virtual disk must have a unique serial number defined directly in the VM specification. For new deployments, explicitly configuring the serial number is recommended, although existing configurations using unique disk names remain valid.

1.2 Serial Number Requirements

• Unique Serial Numbers: Every disk must have a unique serial number cluster-wide.

• Consistent Shared Disk Serials: Shared disks attached to multiple VMs must maintain identical serial numbers on all VM instances.

• Consistent Naming: Disk names must be consistent across all cluster nodes to simplify management and troubleshooting.

• UUID Format: Use UUID format for serial numbers to ensure global uniqueness.

Single VM Disk Configuration Example:

spec:
  template:
    spec:
      domain:
        devices:
          disks:
            - disk:
                bus: scsi
              name: shared-disk-1
              serial: a346eda2-4515-4f67-915e-2ad72617353c
              shareable: true
            - disk:
                bus: scsi
              name: shared-disk-2
              serial: 0a338b98-0351-4986-a589-e55d5d7de4d4
              shareable: true

Multi-VM Shared Disk Configuration Example:

VM1 Configuration:

disks:
  - disk:
      bus: scsi
    name: cluster-shared-disk-1
    serial: 12345678-1234-1234-1234-123456789abc
    shareable: true
 

VM2 Configuration:

disks:
  - disk:
      bus: scsi
    name: cluster-shared-disk-1
    serial: 12345678-1234-1234-1234-123456789abc
    shareable: true
 

1.4 Implementation Best Practices

• Generate unique UUIDs for each disk using standard UUID generation tools.

• Ensure shared disks have the same serial number across all accessing VMs.

• Validate serial number uniqueness and consistency cluster-wide before deployment.

1.5 Clearing Clone Flag

If disks are detected with a clone flag from the explorer, clear the flag by following How to remove the clone_disk flag from a device in vxdisk list to avoid issues.

2.1 Shared Diskgroup Deployment (Non-IOShip)

• Uses OpenShift Virtualization’s default network settings.

• Standard MTU sizes (e.g., 1500 bytes) are supported without additional configuration.

• No specific network tuning is required.

2.2 FSS (IO Shipping) Deployments

• Jumbo Frames must be enabled on all network components with a minimum MTU of 9000 bytes.

• Standard MTU of 1500 bytes is unsupported and will cause performance degradation or operational failures.

• Failure to configure Jumbo Frames properly may severely impact cluster availability and performance.

3.1 Coordination Point Server (CPS) Fencing

• Server-based I/O fencing is supported using Coordination Point Servers (CPS) for deterministic cluster fencing.

3.2 iSCSI Disk-Based Fencing

• Disk-based fencing can be implemented using iSCSI devices visible to the OpenShift VMs as coordination disks.

4. Pre-Deployment Checklist

Before going live, ensure:

• Distinct disks are assigned unique serial numbers.

• Shared disks have identical serial numbers across all VMs accessing them.

• Disk names are consistent across the cluster.

• Network MTU settings are configured appropriately for your deployment (standard MTU for shared diskgroup, Jumbo Frames for IO shipping).

• Fencing configuration (CPS or iSCSI) is properly set up.

• Jumbo Frames are enabled if using IO shipping (FSS).

Adhering to these guidelines will ensure a stable, high-performance Arctera InfoScale environment in OpenShift VM deployments.