Home > Storage > PowerScale (Isilon) > Product Documentation > Storage (general) > Dell PowerScale: Considerations and Best Practices for Large Clusters > Understanding and analyzing group membership
Group membership is one of the key troubleshooting tools for large clusters, where the group composition may change fairly frequently as drives and other components degrade and are SmartFailed out. As such, an understanding of OneFS group membership reporting allows you to determine the current health of a cluster. It also enables a cluster's history to be reconstructed when triaging and troubleshooting issues that involve cluster stability, network health, and data integrity.
Under normal operating conditions, every node and its requisite disks are part of the current group. This can be viewed by running the ‘sysctl efs.gmp.group’ CLI command from any healthy node of the cluster. A OneFS group consists of two parts:
The membership list shows the group members within brackets. Consider the following example:
{ 1-3:0-11 }
This represents a healthy three node cluster, with Node IDs 1 through 3. Each node contains 12 hard drives, numbered zero through 11.
Array IDs differ from Logical Node Numbers (LNNs), the node numbers that occur within node names, and displayed by isi stat. These numbers may also be retrieved on a node using the isi_nodes command. LNNs may be re-used, whereas Array IDs are never re-used. Drive IDs are also never recycled. When a drive is failed, the cluster will identify the replacement drive with the next unused number. However, unlike Array IDs, Drive IDs start at 0 rather than at 1.
Group messages also compact sequential lists into a pair of numbers separated by dashes, as in our previous example of { 1-3:0-11 }. Without dashes, deployed to compress the node list, this group might be listed as:
{ 1:0-11, 2:0-11, 3:0-11 }
When drive 2 is removed from node 2, this consolidated list would become:
{ 1:0-11, 2:0-1,3-11, 3:0-11 }
When a replacement disk is added to node 2, the list would become:
{ 1:0-11, 2:0-1,3-12, 3:0-11 }.
Over time and as the node count grows, these changes can make cluster groups considerably more challenging to read.