Narsing Shah

Creating and Managing Datastores and Raw Device Mappings (RDM) on VMware vSphere

8/15/2016

A Practical Guide for VMware Administrators and Infrastructure Engineers (2016)

Introduction

Storage is one of the most critical components of any VMware vSphere environment. Every virtual machine ultimately depends on the underlying storage subsystem for availability, performance, and scalability.

While VMware VMFS datastores are the standard method for hosting virtual machines, certain enterprise applications require direct access to storage through Raw Device Mapping (RDM).

Creating a VMFS Datastore

Figure 1: Creating a new VMFS datastore from an available SAN LUN in the vSphere Client. Image credit : Gemini.

This article provides a practical walkthrough of creating and managing VMFS datastores and configuring Raw Device Mappings in VMware vSphere. The focus is on production administration rather than theory, covering the tasks commonly performed by VMware administrators in enterprise environments.

1. Storage Planning Before Deployment

Proper planning avoids costly storage migrations and downtime later.

Gather Application Requirements

Before requesting storage from the SAN team, identify:

  • Capacity requirements
  • Expected annual growth
  • Read/write workload
  • Backup requirements
  • Availability requirements
  • Clustering requirements

Example:

Application Initial Size Expected Growth Storage Type
Windows File Server 500 GB 20% yearly VMFS
SQL Database 2 TB High RDM
Oracle Database 5 TB High RDM
Web Server 100 GB Low VMFS

Decide Between VMFS and RDM

Use VMFS when

  • Hosting standard virtual machines
  • Using VMware snapshots
  • Using Storage vMotion
  • Simplifying administration

Use RDM when

  • Microsoft Failover Clustering (MSCS)
  • Large databases
  • Storage-array replication
  • SAN-based backup solutions
  • Applications requiring direct LUN access

Capacity Planning

Avoid filling datastores beyond 80%.

Datastore Capacity : 5 TB
Target Usage       : 4 TB
Reserved Space     : 1 TB

This allows room for VM growth and temporary operations such as snapshots.

Naming Convention

DS_PROD_SQL01
DS_PROD_WEB01
DS_DEV_GENERAL
DS_TEST_VMFS01

Consistent naming simplifies administration and troubleshooting.

2. Preparing Storage on the SAN

Most VMware environments rely on Fibre Channel or iSCSI storage arrays.

Typical SAN administrator tasks include:

  • Create the LUN
  • Assign the LUN ID
  • Present the LUN to ESXi hosts
  • Configure LUN masking
  • Configure Fibre Channel zoning (if applicable)

Verify Connectivity

Navigate to:

Configuration
→ Storage Adapters
→ Rescan All

Or from the ESXi shell:

esxcli storage core adapter rescan --all

Verify New Devices

esxcli storage core device list

Example:

naa.60060160xxxxxxxxxxxxxxxx

Size: 2048 GB
Status: Active

Verify Multipathing

esxcli storage nmp device list

Example:

Path 1   Active
Path 2   Active
Path 3   Standby
Path 4   Standby

3. Creating and Managing VMFS Datastores

Once the storage is visible, create the datastore.

Step 1

Open:

vSphere Client

Configuration

Storage

Add Storage

Select:

Disk/LUN

Step 2

Choose the new LUN.

LUN 20

2 TB

Step 3

Select the filesystem.

VMFS-5

Step 4

Assign the datastore name.

DS_PROD_SQL01

Step 5

Allocate the full available capacity.

Complete the wizard.

The datastore becomes immediately available.

Verify

Check:

  • Capacity
  • Free Space
  • VMFS Version

Extending Existing Datastores

If the SAN administrator expands the LUN:

  1. Rescan storage.
  2. Select the datastore.
  3. Click Increase Capacity.

VMFS expands online without interrupting running virtual machines.

Monitor Usage

Regularly monitor:

  • Capacity
  • Free Space
  • Latency
  • Number of virtual machines

Avoid placing too many high-I/O virtual machines on the same datastore.

4. Configuring Raw Device Mapping (RDM)

RDM provides direct access to a SAN LUN while allowing VMware to manage the virtual machine.

Compatibility Modes

Physical Compatibility Mode

Recommended for:

  • Microsoft Cluster Services
  • SAN replication
  • Storage-array snapshots

Limitations:

  • VMware snapshots are not supported.

Virtual Compatibility Mode

Supports:

  • VMware snapshots
  • Limited Storage vMotion
  • VMware management features

Suitable for most enterprise database workloads.

Creating an RDM

Edit the VM settings.

Select:

Add

Choose:

Hard Disk

Raw Device Mapping

Select the available LUN.

Choose either:

Physical

or

Virtual

Store the mapping file on a VMFS datastore.

Note

Only the small mapping file resides on VMFS. The actual data remains on the physical SAN LUN.

Windows Verification

Open:

Disk Management

The disk initially appears as:

Offline

Bring the disk online, initialize it, and format if required.

Linux Verification

fdisk -l

or

lsblk

Confirm the new disk is visible.

5. Managing Storage Growth and Day-2 Operations

Storage requirements continually increase in production environments.

Expanding Datastore Capacity

Typical workflow:

  1. Expand the LUN on the SAN.
  2. Rescan storage adapters.
  3. Expand the datastore.

No downtime is required.

Expanding Guest OS Disks

Windows:

Disk Management

Extend Volume

Linux:

growpart

resize2fs

xfs_growfs

Storage vMotion

Move virtual machines without downtime.

Right Click VM

Migrate

Change Datastore

Common use cases include:

  • Storage maintenance
  • Datastore balancing
  • Storage hardware refresh

Capacity Monitoring

Monitor:

  • Used Capacity
  • Free Capacity
  • Datastore Latency
  • Queue Length

Review these metrics weekly in production.

6. Performance Optimization and Common Bottlenecks

Storage issues often present themselves as virtual machine performance problems.

Multipathing

Verify the storage path policy.

Round Robin

Example:

Path Selection Policy

Round Robin

Round Robin distributes I/O across multiple storage paths.

Queue Depth

Watch for:

  • High latency
  • Slow virtual machines
  • Database delays

Coordinate queue depth tuning with the storage team.

Datastore Distribution

Avoid placing the following workloads on the same datastore:

  • SQL Server
  • Exchange
  • Oracle
  • File Servers

Distribute workloads across multiple datastores.

Storage Latency Guidelines

Below 10 ms     Excellent

10–20 ms        Acceptable

Above 20 ms     Investigate

Above 50 ms     Critical

Snapshot Best Practices

Delete snapshots immediately after backups or maintenance.

Never leave snapshots active for extended periods.

7. Troubleshooting Storage Issues

Datastore Not Visible

Verify:

  • Storage presentation
  • Adapter rescan
  • HBA health
  • SAN zoning

Rescan storage:

esxcli storage core adapter rescan --all

LUN Missing

esxcli storage core device list

If no device appears, verify:

  • SAN presentation
  • LUN masking
  • Fibre Channel zoning
  • iSCSI connectivity

RDM Cannot Be Created

Possible causes:

  • LUN already assigned
  • Incorrect presentation
  • Unsupported compatibility mode
  • Existing partitions

Path Failure

esxcli storage core path list

Look for:

Dead

Disabled

Standby

Healthy environments should always have redundant active paths.

APD (All Paths Down)

Symptoms include:

  • VM freezes
  • Datastore unavailable
  • Storage timeout messages

Investigate:

  • SAN switches
  • HBAs
  • Storage processors
  • Cabling

PDL (Permanent Device Loss)

Possible causes:

  • Storage unpresentation
  • SAN changes
  • Storage controller failures

Useful ESXi Commands

esxcli storage core adapter list
esxcli storage core device list
esxcli storage core path list
esxcli storage filesystem list

8. Enterprise Best Practices

  • Standardize datastore naming.
  • Separate high-I/O workloads.
  • Maintain at least 20% free datastore capacity.
  • Verify multipathing on every ESXi host.
  • Monitor storage health proactively.
  • Remove snapshots promptly.
  • Document every storage-related change.
  • Perform periodic storage audits.

Conclusion

VMware vSphere provides two primary methods of presenting storage to virtual machines: VMFS datastores and Raw Device Mapping (RDM).

VMFS remains the preferred choice for most workloads because of its flexibility, ease of management, and support for VMware features such as snapshots and Storage vMotion. RDM should be reserved for specialized enterprise workloads that require direct SAN access or integration with storage-array capabilities.

By following the planning guidelines, implementation procedures, troubleshooting techniques, and operational best practices presented in this guide, VMware administrators can build a scalable, resilient, and high-performing storage infrastructure capable of supporting enterprise virtualization workloads.