Following on from Dan's how-to regarding the configuration of SunCluster 3.2 within VMware Fusion, here’s my additional notes on doing this from a minimal install (i.e core software group) so as to keep the size of the VMs down.
First thing you’ll want to do is install some packages to make your life a bit less painful. Connect the DVD device with the Solaris 10 ISO and mount it within your VM:
# mount -F hsfs -o ro /dev/dsk/c0d0t0s0 /mnt/cdrom
Then we can install some packages from /mnt/cdrom/Solaris_10/Product. Let’s start off with some basic utilities, bash, and online documentation (man pages):
# yes | pkgadd -d . SUNWxcu4 # yes | pkgadd -d . SUNWadmfr SUNWadmfw # yes | pkgadd -d . SUNWbash # yes | pkgadd -d . SUNWdoc SUNWman
SSH is also quite useful:
# yes | pkgadd -d . SUNWsshr SUNWsshu SUNWsshdr SUNWsshdu SUNWsshcu # /lib/svc/method/sshd -c # svcadm enable ssh
The second command will create your RSA and DSA host keys, without this you won’t be able to connect.
Then there’s a bunch of additional prerequisite packages you need to install before SunCluster itself. These are as follows:
SUNWtcatu SUNWj3rt SUNWj3dev SUNWmfrun SUNWxwrtl SUNWxwice SUNWxwplt SUNWctpls SUNWxwfnt SUNWxwplr
You also need some Solaris Zones related gubbins otherwise you’ll see various errors pertaining to missing libraries such as libzonecfg.so.1:
SUNWzoneu SUNWzoner SUNWpool SUNWluu SUNWluzone SUNWlur SUNWpoolr SUNWlucfg
Finally, the core / reduced networking group installation also disables some RPC related services that are necessary for cluster node communication. You’ll find that installation hangs, for example, when the primary node is waiting for the secondary to reboot. To remedy this, do the following:
# svccfg -s rpc/bind setprop config/local_only = false # svccfg -s rpc/bind setprop config/enable_tcpwrappers = false # svccfg -s rpc/bind listprop # svcadm refresh rpc/bind
One other oddity I noticed was that I was seeing a bunch of RPC timeouts when looking at anything IPMP related. This turned out to be some DNS resolution going on as a result of the IP addresses assigned to the cluster interconnects and my ISP returning bullshit records. It’s easily fixed by editing /etc/nsswitch.conf and amending the ipnodes line as follows:
You should then be ready to kick off the install of SunCluster, and everything else is exactly as per Dan’s guide.