Veritas Volume Manager (VxVM) 7.4.2.x can now import Hardware Replicated disks that are in a SPLIT state using vxdg -c -o usereplicatedev=only

Problem

Traditionally, Hardware Replicated devices such as EMC SRDF, Hitachi TrueCopy could not be imported when in a "SPLIT" state.

The "SPLIT" state means both the source & target disks are available in a read-write (RW) state on the respective hosts. 

Whilst "SPLIT" the Hardware Replication is "not ready" and no data is being transferred between the source & target disks.

Veritas Volume Manager (VxVM) would fail the disk group import with "Device is a hardware mirror", as it believes the disks are Hardware Clones (like EMC BCV, SYMCLONE, Hitachi ShadowImage and so on).

Error Message

Example: EMC SRDF

1.] Even though the EMC SRDF-R2 (target) disks have been "SPLIT" using EMC SYMCLI command:

# symrdf -g EMC Device Group split

2.] DMP & VxVM extended attributes refreshed

# vxdisk scandisks

3.] The VxVM disk group (i.e. SRDFDG) operation still fails for the EMC SRDF-R2 "SPLIT" disks:

# vxdg -c import SRDFDG
VxVM vxdg ERROR V-5-1-19179 Disk group SRDFDG: import failed:
Device is a hardware mirror

Cause

VxVM in connection with Veritas Dynamic Multi-pathing (DMP) use a series of extended attributes to determine the type of LUN presented to the host.

Sample syntax

# vxdisk -px LIST_CLONE list | egrep 'DEVICE|emc'
DEVICE                   DG_NAME   REPLICATED REPLICATED_TYPE HARDWARE_MIRROR LUN_SIZE   NR_DEVICE DEVICE_TYPE DDL_DEVICE_ATTR
emc0_033c                LINUXSRDF yes        rep_sec         yes             20974080   no        -           WD,srdf-r2,tprclm
emc0_033d                LINUXSRDF yes        rep_sec         yes             20974080   no        -           WD,srdf-r2,tprclm
emc0_033e                LINUXSRDF yes        rep_sec         yes             20974080   no        -           WD,srdf-r2,tprclm
emc0_033f                LINUXSRDF yes        rep_sec         yes             20974080   no        -           WD,srdf-r2,tprclm
emc0_035e                LINUXSRDF yes        rep_sec         yes             20974080   no        -           WD,srdf-r2,tprclm
emc0_035f                LINUXSRDF yes        rep_sec         yes             20974080   no        -           WD,srdf-r2,tprclm
emc0_0264                LINUXSRDF -          -               no              20974080   no        -           tprclm
emc0_0265                LINUXSRDF -          -               no              20974080   no        -           tprclm
emc0_0266                LINUXSRDF -          -               no              20974080   no        -           tprclm
emc0_0267                LINUXSRDF -          -               no              20974080   no        -           tprclm
emc0_0268                LINUXSRDF -          -               no              20974080   no        -           tprclm
emc0_0269                LINUXSRDF -          -               no              20974080   no        -           tprclm
 

In the above example, we can see EMC SRDF-R2 devices in a "Write-Disabled" (WD) state.

The Storage Array Vendors cannot expose all attributes via SCSI inquiries.

At this time, Storage Vendors currently don’t expose the "SPLIT" state, so VxVM is unaware of what has occurred behind-the-scenes.

Solution

The vxdg –c import option was introduced to handle Hardware Replicated devices (such as EMC SRDF, Hitachi TrueCopy etc).

VxVM has introduced NEW improved logic during the LUN selection process when using DMP.

NEW vxdg "-o usereplicatedev=only" CLI option must be used in connection with “-c” for H/W Replicated disks only.

This is also required for EMC SRDF Metro Mirror type devices which appear as srdf-r1 disks.

# vxdg -cs import LINUXSRDF
VxVM vxdg ERROR V-5-1-0 Disk group LINUXSRDF: import failed:
Replicated dg record is found.
Did you want to import hardware replicated LUNs?
Try vxdg [-o usereplicatedev=only] import option with -c[s]
 

The enhanced vxdg import syntax logic will explicitly cater for Hardware Replicated devices and for those devices that are in a SPLIT state. This functionality is restricted to Veritas DMP only

The command vxdg -c [s] -o usereplicatedev=only will exclusively select Hardware Replicated devices only.
 

The enhanced vxdg -c -o usereplicatedev=only option reduces the chances of accidentally importing the incorrect disk group image.

Requires DMP & VxVM attributes to be refreshed prior to importing using "vxdisk scandisks".

Figure 1.0

Fencing enabled:

# vxdg -Ccs -t -o usereplicatedev=only -o groupreserve="VCS" import LINUXSRDF

Non-Fencing:

# vxdg -Ccs -t -o usereplicatedev=only  import LINUXSRDF
 

NOTE: The "-s" option is required for Shared disk groups in connection with Clustered Volume Manager (CVM). The "-C" option is required to clear the stale hostid from the source system.
 

Veritas has released a private hot-fix for InfoScale on Linux (Veritas Volume Manager 7.4.2 Hot Fix 2902). The product enhancement will be ported to other platforms, Solaris & AIX very soon.

Patch ID: 7.4.2.2902

* 4066786 (Tracking ID: 4066785)

SYMPTOM:
When the replicated disks are in SPLIT mode, importing its disk group failed with "Device is a hardware mirror".

DESCRIPTION:
When the replicated disks are in SPLIT mode, which are readable and writable, importing its disk group failed with "Device is a hardware mirror". Third party doesn't expose disk attribute to show when it's in SPLIT mode. With this new enhancement, the replicated disk group can be imported with option -o usereplicatedev=only.

RESOLUTION:
The code is enhanced to import the replicated disk group with option -o usereplicatedev=only.
 

Please contact Veritas Technical Support if you require this private hot-fix.