Business Continuance Volume (BCV) devices remain in error state after a BCV split. scsimgr lun_map shows the path State as FAILED.

book

Article ID: 100008849

calendar_today

Updated On:

Description

Error Message

# vxdisk list
DEVICE       TYPE            DISK         GROUP        STATUS
disk52       auto            -            -            error
disk53       auto            -            -            error
# scsimgr lun_map -D /dev/rdisk/disk53
 
       LUN PATH INFORMATION FOR LUN : /dev/rdisk/disk53
 
Total number of LUN paths     = 1
World Wide Identifier(WWID)    =
 
LUN path : lunpath17
Class                         = lunpath
Instance                      = 17
Hardware path                 = 42/0/0/2/0/0/0/4/0/0/1.0x5006048ad5f116c0.0x4023000000000000
SCSI transport protocol       = fibre_channel
State                         = FAILED
Last Open or Close state      = DISABLE_OPENS 

Cause

The issue is seen when the system is booted with BCV  devices in an established state (write-disabled). DMP opens the path of the device in read only mode and caches the open. Any subsequent application open in write mode fails. vxdmpadm dump command shows that an open count is maintained for the device.

When the mode of BCV devices are changed from write protected to read-write mode (after a BCV split operation), the OS is unable to close the device and re-open in the correct mode (because DMP keeps an opencount on the device).

When the device was removed from Volume manager control with "vxdisk rm " command and rescanned with "vxdctl enable" command, DMP relinquished the open on the device and the OS was able to close and re-open the device in the correct mode.

 

This issue is only seen when dmp_cache_open is set to ON and BCV  or SRDF  operation is performed on the devices.

dmp_cache_open tunable information:

dmp_cache_open

                 If set to on, the  first  open on  a device that is performed by an  array  support  library (ASL)is cached. This enhances  the  performance  of device discovery by minimizing the overhead caused by  subsequent opens by ASLs. If set to off, caching is not performed. The default setting is on.

This tunable can help with improved device discovery times (vxdisk scandisks, vxdctl enable) or diskgroup imports (vxdg import) in environments where there are large number of disks and also the caching prevents accidental Dynamic Reconfiguration of the devices that are under VxVM control.

To check the current setting:
# vxdmpadm gettune dmp_cache_open
Tunable                         Current Value Default Value
------------------------------  -------------         -------------
dmp_cache_open         on                    on

Resolution

Some arrays provide a way to change the properties of LUNs. For example, the EMC symmetrix array allows write-protected (Read-only), and read-write enabled LUNs. Before changing the properties of a LUN, you must remove the device from Veritas Volume Manager (VxVM) control.

Steps to change the properties of a LUN

1.) If the device is part of a disk group, move the disk out of the disk group.
# vxdg -g dgname rmdisk daname

2.) Remove the disk from the vxdisk list. In a cluster, perform this step from all of the nodes.
# vxdisk rm da-name

3.) Change the device characteristics.

4.) Use DMP to perform a device scan. In a cluster, perform this command on all the nodes.
# vxdisk scandisks

5.) Add the device back to the disk group.
# vxdg -g dgname adddisk daname


Applies To

  • HP-UX  11.31 IA architecture
  • VxVM  5.1SP1
  • dmp_cache_open tunable set to ON (which is the default option)
  • BCV  or SRDF  operation is performed on the host

Issue/Introduction

When the BCV  or SRDF (Symmetrix remote data facility) devices change from write-disabled to write-enabled state, Volume Manager fails to recognize this change. Devices remain in error state. HP-UX  sees the device paths in a FAILED state.