Problem and solution when the round robin load balancing policy of NMP does not take effect in HP_UX

You can perform the following operation when round robin load balancing policy of NMP does not take effect in HP_UX as follows:
1. Issue Description
What can I do when the round robin load balancing policy of NMP does not take effect in HP_UX?
Situation: HP 11i v3 is delivered with the Native Multi-Path multipathing software and the default load balancing policy round robin. Multiple paths from the hosts to controller A or controller B on the storage array do not take effect, and you can only deliver I/Os through one path.
2. Solution
a. Enter scsimgr get_attr -a leg_mpath_enable on the host to check whether the current system has enabled NMP multipathing software.
b. Enter scsimgr get_attr -D /dev/rdisk/diskXX -a leg_mpath_enable to check whether LUN has enabled the NMP software.
c. Enter scsimgr get_attr -D /dev/rdisk/diskXX to check whether LUN has a default load balancing policy round robin.
d. Enter scsimgr lun_map -D /dev/rdisk/diskxx to check which LUN has active status of all paths.
e. Enter scsimgr get_stat -H Hardware path to check the I/O statistics of each path to the LUN. If the I/Os of one path is small, nearly no I/O is delivered on this path.
f. Above all, although the NMP multipath configuration does not cause any failure, there is a problem which is referred on released notes.
You solve the problem by setting alua_enabled of a LUN to false. However, based on lab tests, this method can deliver I/Os through multiple paths but the performance of this method degrades compared to I/Os delivered through one path. You are advised to use one path not to use multiple paths.

Scroll to top