Power9 management server

Power9 management server

  • Only a single instance of all management services is supported (that is, a single management server per deployment).
  • All nodes in the storage cluster must run the same IBM Storage Scale System release and IBM Storage Scale version to ensure compatibility and stability.
  • It is recommended to upgrade to the latest available version before adding a new building block to the cluster.

Enabling remote management

To enable remote support in data centers with restrictive access, the following points must be considered:
  • Connect HMC2 ports on all Power servers (IBM Storage Scale System 5000 or Power9 management servers) to the campus network and assign a campus IP address. This connection enables remote recovery and debugging during outages.
  • Connect C11-T3 ports on Power9 nodes to the campus network to support remote access.
  • Enable campus connectivity on IBM Storage Scale System 3200 by consulting IBM support about by using a USB-to-Ethernet dongle.
  • Connect a free network port on each IBM Storage Scale System 3000 canister to the campus network.

    Also, install SMART PDUs in each 3000 frame to support remote power cycling.

  • Connect Utility Nodes (x86-based) to the campus network. Although this is an optional task, this is recommended to improve remote support and management capabilities.

Preparing a stretch cluster

Consider A and B sites of the same cluster.
  1. Ensure all management networks (FSP, BMC, MGMT) and the high-speed network (InfiniBand or Ethernet) are flat and operate within a single namespace.
  2. Ensure each node has a unique hostid. If not, assign a unique IP address and hostname hostid to each node for both the management and high-speed networks.
    • Check hostid values manually.
      mmdsh -N <NodeClass or NodeList> hostid
    • If duplicates are found, generate a new hostid:
      genhostid
  3. Verify that all nodes are reachable from one another over all networks.
  4. Define a common /etc/hosts file or configure DNS appropriately, and distribute it to all nodes by using IBM Storage Scale System deployment tools.
  5. Run gnrhealthcheck before you create recovery groups in stretch clusters.