There seems to be a locking problem in a dual 2100 dssi
cluster running Rdb that resides on shared DSSI disks.
The Dlock manager takes A LOT of time to grant a lock on
the node that masters the lock tree.(Lockdirwt either 1
in both nodes or 1 and 2). The alloclass is set to 1 in
one node and the HSD30 dssi and to 2 in the second node.
I understand that both systems access the dssi via the
SHAC dssi connection.
Can it happen that the Distributed lock manager tries
to remaster the lock tree using the alloclass as is the
dssi were not present, for instance exchanging locking
information via the Ethernet?