EMC is aware of this design and they have documented the behaviour under their Primus(R) eServer solution (emc115907).

The "legacy" workaround of placing the BCV luns in "Defined" state whilst the BCV luns are in an established state, and made "Available" again once the luns are "split" is no longer required with 6.0.5 and 6.1.0.100. 

The EMC utility "rmbcv" traditionally marked the specified BCV lun(s) in a "Defined" state whilst in an established state.
The "mkbcv" utility was then responsible for making the specified BCV lun(s) transition from the "Defined" to "Available" state once the BCV "split" operation had completed for the corresponding BCV devices.

The Veritas updated process to handle EMC BCV devices with 6.0.5 and 6.1.0.100 onwards on AIX is as follows:

1.] Following a system reboot/restart, the EMC BCV devices must be manually defined as "Available" using the EMC script "/usr/lpp/EMC/Symmetrix/bin/mkbcv -a"

Sample output:

# lsdev -Cc disk | grep TimeFinder

hdisk4   Defined   07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk5   Defined   07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk6   Defined   07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk7   Defined   07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk8   Defined   07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk9   Defined   07-09-02     EMC Symmetrix FCP Disk TimeFinder
.
.

# /usr/lpp/EMC/Symmetrix/bin/mkbcv -a

hdisk4 Available

hdisk5 Available

hdisk6 Available

hdisk7 Available

hdisk8 Available

hdisk9 Available
.
.



2.] Using the AIX CLI command "lsdev -Cc disk", confirm the EMC devices are now shown as "Available".

Sample output:

# lsdev -Cc disk | grep TimeFinder

hdisk4   Available 07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk5   Available 07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk6   Available 07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk7   Available 07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk8   Available 07-09-02     EMC Symmetrix FCP Disk TimeFinder
hdisk9   Available 07-09-02     EMC Symmetrix FCP Disk TimeFinder
.
.

3.] In order for VxVM to detect the OS device state change, "vxdisk scandisks" needs to be executed.

# vxdisk scandisks

# vxdisk -eo alldgs list | grep -i bcv

emc1_00fe    auto:aixdisk   -            -           online udid_mismatch hdisk45          lun bcv
emc1_00ff    auto:aixdisk   -            -           online udid_mismatch hdisk10          lun bcv
emc1_0100    auto           -            -           error                hdisk30          lun bcv nr WD
emc1_0101    auto           -            -           error                hdisk48          lun bcv nr WD
emc1_0122    auto           -            -           error                hdisk49          lun bcv nr WD
emc1_0123    auto           -            -           error                hdisk14          lun bcv nr WD


Veritas DMP is capable of exposing multiple device characteristics for EMC (Symmetrix) related devices, known as "extended attributes".
In the above instance, DMP is reporting the EMC LUN type "bcv" and device state "nr".

DMP extended attribute content is intentionally suppressed for other Third-Party mult-pathing drivers (TPD).


Note: The "mkbcv -a" process need to be manually repeated each time the server is restarted.

 

 

Applies To

AIX

EMC TimeFinder (BCV) devices on AIX in connection with Veritas Volume Manager (VxVM) are intentionally not seen until the device state is manually changed from "Defined" to "Available". During the system restart process EMC TimeFinder (BCV) devices will be created by "cfgmgr" in a "Defined" state to avoid PVID conflicts in relation to the source and clone (BCV) devices.

The EMC BCV device state can be manually adjusted to report an "Available" or "Defined" status by the use of two EMC scripts "mkbcv" and "rmbcv" respectively. Veritas has previously recommended offlining the clone (BCV) devices and placing them in a "Defined" state (using the "rmbcv -a" script) to avoid VxVM and other command related hangs, whilst the EMC BCV devices are in a Not-Ready (NR) state.

The AIX device state can be validated using the "lsdev -Cc disk" CLI command.

With VxVM 6.0.5 and 6.1.0.100 onwards, once the server has been restarted the BCV devices can now be made available using the EMC "mkbcv -a" script regardless of the BCV state.
VxVM fix 6.0.5.003 includes an enhancement to suppress the DMP related message "VxVM vxdmp V-5-3-0 dmp_indirect_ioctl: Open failed on path <major/minor> with ret 47" for devices in a Write-Disabled (WD) or Not-Ready (NR) state.

Note: The additional enhancement will be ported to 6.1.x (not included in 6.1.1) and 6.2 release trains for AIX.

The offlining and removal of the BCV devices was done originally to ensure that there were no contention issues during the resychronization process between the source and clone (BCV) devices.