Not sure what "-a" means, but the process should look the same on each node, and as Solaris starts mpathd, this would suggest the 2 nodes are configured differently so you should try to find this difference (have a look at /etc/default/mpathd as a starting point)
The VCS bundled agents guide gives some checks:
Checklist to ensure the proper operation of MultiNICB
For the MultiNICB agent to function properly, you must satisfy each item in the
following list:
â– Each interface must have a unique MAC address.
â– A MultiNICB resource controls all the interfaces on one IP subnet.
â– At boot time, you must configure and connect all the interfaces that are under
the MultiNICB resource and give them base IP addresses.
â– All base IP addresses for the MultiNICB resource must belong to the same
subnet as the virtual IP address.
â– Reserve the base IP addresses, which the agent uses to test the link status, for
use by the agent. These IP addresses do not get failed over.
â– The IgnoreLinkStatus attribute is set to 1 (default) when using trunked
interfaces.
â– If you specify the NetworkHosts attribute, then that host must be on the same
subnet as the base IP addresses for the MultiNICB resource.
â– Test IP addresses have "nofailover" and "deprecated" flags set at boot time.
â– /etc/default/mpathd has TRACK_INTERFACES_ONLY_WITH_GROUPS=yes.
â– If you are not using Solaris in.mpathd, all MultiNICB resources on the system
have the UseMpathd attribute set to 0 (default). You cannot run in.mpathd on
this system.
â– If you are using Solaris in.mpathd, all MultiNICB resources on the system have
the UseMpathd attribute set to 1.
Mike