MPIO fail to failover MPIODISK - HELP!

[Tony Chua]

Hello All,

DL380G4
Windows 2003 Ent SP2
MPIO V2.00.02
SCSIport 9.1.0.13
HBA FC2214 x2 BIOS 1.45 FW3.03.18
EVA8000

When I disable one of SAN switch port, MPIO fail to failover MPIODISK...and the cluster failover! Event messages list below for reference. please kindly help...

An unrecoverable path failure occurred on SCSI address Port 3 B0 T0 L1. Disk 600508B400106B250000E00000DE0000 is still accessible over redundant path(s).

A fail-over on \Device\MPIODisk0 was attempted, however the attempt failed. The devices will be removed.

The device 'HP StorageWorks EVA 8000 Storage Array Controller Device' (SCSI\Array&Ven_HP&Prod_HSV210&Rev_5000\6&2bf64f5f&0&000) disappeared from the system without first being prepared for removal.

he DSM has completed remove processing for path (SCSI address Port 3 B0 T0 L1) to multipath capable disk 600508B400106B250000E00000DE0000.

The device 'HP StorageWorks EVA 8000 Storage Array Controller Device' (SCSI\Array&Ven_HP&Prod_HSV210&Rev_5000\6&2bf64f5f&0&010) disappeared from the system without first being prepared for removal.

An unrecoverable path failure occurred on SCSI address Port 3 B0 T1 L1. Disk 600508B400106B250000E00000DE0000 failed due to no redundant paths available.

[Víctor Cespón]

How do you have your SAN configured? Each server with 2 FC ports? Each connected to a fabric? Ports on the EVA connected to both fabrics too?

The messages above indicate that the server loses connection to BOTH controllers of the EVA8000, so it loses total access to the disk (that triggers the cluster failover).

Normally, how many "EVA 8000 Storage Array Controller" do you see in Windows device manager?

[Tony Chua]

Yes, each server have 2 fc port to 2 san switch and the server see 4 EVA8000 controller (2 for each path)....

[Víctor Cespón]

That means you are not using all ports on EVA8000, each controller has 4 external ports.
You need to check the logs of the SAN switches that interconnect that. A failure on a switch will make two of the EVA8000 devices dissapear, like the log shows.
But then you get that a disk "failed due to no redundant paths available". It should still be available through the other fabric.

[shIVinator.]

You might also want to try to update MPIO. HP had some issues with clustering and MPIO pre version 2.01.00 (2.01.01 is the latest).

2.01.xx also has added support for clustering.


Here are the release notes:
+++++++++++++++++++++++++++++++++++++++++++++
Version 2.01.00
* Added support for Adaptive Load Balance (ALB) algorithms that utilize the array specific features to deliver better performance for host I/O requests.
* Fixed a possible occurrence of system crash (stop code 0x35) while using load balancing feature under cluster environments.
* Updated to MPIO framework 1.17.

NOTE: HP recommends that you upgrade to HP MPIO EVA DSM 2.01.00 only if you want any of the above features.


Version 2.00.02
* Fixed the occurrence of I/O failures observed during controller re-synchronization operations (as a result of, but not limited to, code load and controller restarts) for EVA XCS 6.x.

NOTE: HP recommends that you install HP MPIO EVA DSM 2.00.02 if you are using a SAN configuration, comprising windows host(s) connected to multiple EVA 4000/6000/8000 disk arrays with a combination of both XCS 5.x and XCS 6.x.
+++++++++++++++++++++++++++++++++++++++++++++