Installation: reference

This topic provides information on adding IBM Spectrum Scale™ nodes to an ESS cluster and node name considerations.

Adding IBM Spectrum Scale nodes to an ESS cluster

IBM Spectrum Scale node configuration is optimized for running IBM Spectrum Scale RAID functions.

ESS cluster node configuration is optimized for running IBM Spectrum Scale RAID functions. Protocols, other gateways, or any other non-ESS services must not be run on ESS management server nodes or I/O server nodes. In a cluster with high IO load, avoid using ESS nodes as cluster manager or filesystem manager. For optimal performance the NSD client nodes accessing ESS nodes should be properly configured. ESS ships with gssClientConfig.sh script located in /usr/lpp/mmfs/samples/gss/ directory. This script can be used to configure the client as follows:
```
/usr/lpp/mmfs/samples/gss/gssClientConfig.sh <Comma Separated list of 
client nodes or nodeclass>
```
You can run the following to see configuration parameter settings without setting them:
```
/usr/lpp/mmfs/samples/gss/gssClientConfig.sh -D
```
After running this script, restart GPFS™ on the affected nodes for the optimized configuration settings to take effect.

Important: Do not run gssClientConfig.sh unless you fully understand the impact of each setting on the customer environment. Make use of the -D option to decide if all or some of the settings might be applied. Then, individually update each client node settings as required.
When IBM Spectrum Scale nodes deployed with protocols are added to the ESS cluster, quorum, cluster manager, and filesystem manager functions should be moved from the ESS to the protocol nodes after adding protocol nodes to the cluster.

For information about adding an IBM Spectrum Scale protocol node to an ESS cluster, see:

Node name considerations

Carefully select the hostname, suffix, and prefix of the management server and I/O server so that the hostname used in the high-speed network and by the ESS cluster can be generated from the suffix or prefix.

High-speed hostnames

Example 1:

a-bcd-edf-1 
a-bcd-edf-2 
a-bcd-edf-3 
a-bcd-edf-4

Here, a-bcd- is the prefix and edf-1, edf-2, edf-3, and edf-4 are the xCAT names of the nodes.

Example 2:

1-a-bcd-edf 
2-b-bcd-edf 
3-c-bcd-edf 
4-d_bcd_edf

Here, -edf is the suffix and 1-a-bcd, 2-a-bcd, 3-a-bcd, and 4-a-bcd are the xCAT names of the nodes.

If possible, avoid using high-speed node names with variations at the beginning and the end, such as:

A-a-bcd-edf-1 
B-b-bdc-edf-2 
C-c-bcd-edf-3 
D-d-bcd-edf-4

In such cases, use the -N option and specify the node list with the gssgencluster and gssgenclusterrgs commands. The node names must be reachable from the management server node. xCAT requires that the target nodes be part of a node group and a warning might be issued if the hostname is not defined as an xCAT object.

Example:

The xCAT hostnames are gssio1, gssio2, gssio3, and gssio4.
The high-speed hostnames are A-test1, B-test2, C-test3, D-test4. These hostnames are reachable from the management server node. They are not defined in xCAT.

Run:

gssgencluster -C test01 -N A-test1,B-test2,C-test3,D-test4