ESXi Multipathing Decoded: MRU, Fixed, and Round Robin

Posted on Updated on

When you present a LUN to an ESXi host, the Native Multipathing (NMP) engine automatically assigns a policy based on the type of storage array detected. However, as an admin, you need to understand why a policy was chosen—and when you should manually intervene.

1. Most Recently Used (MRU)

Best For: Active/Passive Arrays. MRU selects the first working path it finds at boot. If that path fails, it switches to a standby path.

  • Key Behavior: It does not fail back. Even if the original path becomes healthy again, the host stays on the current path. This prevents “path thrashing” on Active/Passive arrays where switching controllers is an expensive operation.

2. Fixed

Best For: Active/Active Arrays. The Fixed policy uses a specific “Preferred Path.” If the preferred path fails, it moves to an alternative.

  • Key Behavior: It does fail back. As soon as that designated preferred path is back online, the host immediately switches back to it.

3. Round Robin (RR)

Best For: Load Balancing (Active/Active or ALUA). Round Robin rotates through all available paths to distribute the I/O load.

  • Active/Active: Uses every available path.
  • Active/Passive: Only uses all paths leading to the active controller.

Note: For Microsoft Failover Clusters (MSCS), Round Robin is only supported on ESXi 5.5 and later.

4. Fixed with Array Preference (FIXED_AP)

Introduced in ESXi 4.1 for ALUA-capable arrays, this policy lets the storage array tell the host which path is the “optimal” one.

  • Note: This was removed in ESXi 5.0 in favor of letting the NMP automatically select MRU or Fixed based on the array’s ALUA response.

⚠️ Critical Warnings for Admins

  1. Don’t Fight the NMP: VMware generally warns against manually changing a LUN from Fixed to MRU. The host chooses the policy based on the hardware it detects; forcing a change can lead to instability.
  2. Verify Vendor Support: Round Robin is powerful but not supported by every array. Always check the VMware Compatibility Guide before making it your default.
  3. MSCS Limitations: If you are virtualizing SQL clusters or other failover clusters, double-check your ESXi version before toggling Round Robin, or you risk losing disk heartbeat connectivity.

#VMware #ESXi #StorageAdmin #vSphere #Multipathing #SysAdmin #ITPro #Virtualization #LazyAdmin #DataCenter #StorageTips

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.