PowerPath: Presenting disks to the host with Powerpath 6.1 in a Microsoft cluster brings all the cluster resources offline for all the disks

Customer upgraded PowerPath to version 6.1 and 6.2 on a Microsoft Failover Cluster.

Host is running PowerPath for Windows 6.1.

Windows 2008 SP2, 2008 R2 SP1, 2012, 2012 R2, and 2016.

Whenever a disk is presented to a Windows server which is a part of a Failover cluster and each disk has more than 8 paths, the cluster resources go temporarily offline.

SYSTEM EVENT LOGS:    

The issue is due to a combination of the following two changes that went into PowerPath for Windows 6.1:   

• Failure of all IO (including  Persistent Reservation In ) that gets a Unit Attention 6/3f/0e.

The change was made so that Disk class driver, above us, can see the Unit Attention and initiate a bus rescan, leading to automatic discovery of new devices. The IO is retried by the Disk class driver after 5 seconds.

• Load balancing of five non-R/W SCSI commands, including Persistent Reservation In.

The change was done to avoid congestion on one path by spreading the most common non R/W commands across different paths.

As a result of bullet 1, PRI commands from the cluster are being failed back to Disk, with a retry interval of 5 seconds.

As a result of bullet 2, each retry of PRI command happens on a different path, leading to a Unit Attention being received on each retry (once for each path).

As a result of both changes, it is taking almost 20 seconds for the clusters PRI command to complete. This delay makes the cluster believe that it has lost access to the disk.

Note: This is a PowerPath for Windows bug and can affect any version of Windows installed.

The issue is addressed in PowerPath 6.2 P01. 

Powerpath for Windows 6.0 SP2 does not report this issue.