In this blog, we will attempt to build Site B cluster using the old method by using mm commands. There are bit more steps in this process than doing it via the Spectrumtool kit but it always good to know the manual steps involved.
We start of by creating node files so open vi editor under /root/nodes-siteb.txt. In here, we use the daemon name directly as that’s lot quicker
- Daemon name (-d): the address GPFS uses between nodes, on the daemon network (VLAN 1614).
- Role: quorum-manager votes and can run the cluster; quorum votes only.
siteb-mgmt01-d:quorum-manager:siteb-mgmt01
siteb-nsd01-d:quorum:siteb-nsd01
siteb-nsd02-d:quorum:siteb-nsd02
Create the cluster
mmcrcluster -N /root/nodes-siteb.txt -C siteb- -N /root/nodes-siteb.txt: who is in the cluster, read from our node file.
- -C siteb: what the cluster is called
Check that the cluster was created
Because it’s a manual install, we do need to accept the EULA otherwise GPFS won’t start so the command mmchlicense declares the nodes as server licensed.
mmchlicense server --accept -N all
Start GPFS automatically at boot
mmchconfig autoload=yes
Start GPFS on all nodes
mmstartup -a
Check the status of GPFS by running
mmgetstate -a
All steps from here is same as the Site A so lets turn off Ubuntu periodic updates on all 3 nodes
Turn off Ubuntu’s periodic updates on all three nodes vi /etc/apt/apt.conf.d/10periodic
GPFS opens extra network connections between nodes for some jobs, such as copying data and running commands. So, we will set in ephemeral port range between 60000 and 61000, so a firewall only needs that one range opened. This is the manual equivalent of the toolkit’s --ephemeral_port_range.
mmchconfig tscCmdPortRange=60000-61000
Put the mm commands in root’s PATH on all 3 nodes
for h in siteb-mgmt01 siteb-nsd01 siteb-nsd02; do ssh $h "echo 'export PATH=\$PATH:/usr/lpp/mmfs/bin' > /etc/profile.d/gpfs.sh"; doneSo that’s about the manual method of configuring GPFS but I still prefer the spectrum scale toolkit