I am now finally able to reboot nodes in my cluster and they aren't stuck at
'waiting for quorum'.
What works correctly is this:
setup the 'public facing' VLANS and IP addresses and one vlan called 'vmltx2'
on the on-motherboard NIC (rge0). Leave the e1000g0 unconfigured, plugged into
an untagged VLAN on the switch. There was some frustrating difficulty with
scinstall NOT correctly running in interactive mode on nodes which had never
had a second Intel NIC installed, but once I got the command line syntax from
ONE node I was able to apply that to all nodes.
My command for the first node (mltstore1) as generated by scinstall interactive
mode:
scinstall -i \
-C mltcluster0 \
-F \
-G lofi \
-T
node=mltstore1,node=mltstore0,node=mltproc0,node=mltproc1,authtype=sys \
-w
netaddr=172.16.0.0,netmask=255.255.240.0,maxnodes=64,maxprivatenets=10,numvirtualclusters=12
\
-A trtype=dlpi,name=e1000g0 -A trtype=dlpi,name=vmltx2 \
-B type=switch,name=switch1 -B type=switch,name=switch2 \
-m endpoint=:e1000g0,endpoint=switch1 \
-m endpoint=:vmltx2,endpoint=switch2 \
-P task=quorum,state=INIT
And my command for all subsequent nodes as generated by scinstall interactive
mode:
scinstall -i \
-C mltcluster0 \
-N mltstore1 \
-G lofi \
-A trtype=dlpi,name=e1000g0 -A trtype=dlpi,name=vmltx2 \
-m endpoint=:e1000g0,endpoint=switch1 \
-m endpoint=:vmltx2,endpoint=switch2
In short: DLPI support appears to be required for the private interconnect, but
is NOT required on the 'first boot when joining the cluster'; further, DLPI is
not required to get TCP/IP connectivity; further, I can't figure out how to get
the 'private interconnect over VLAN' piece working. I don't understand at all
why 3 of 4 nodes would work and reboot correctly when using the 'non-DLPI'
devices. Of course, perhaps I'm completely misunderstanding all of this... but
for the moment it appears that I am able to reboot my nodes and they join the
cluster so I'm going to proceed.
--
This message posted from opensolaris.org
