System state

The state of the system holds all of the configuration and internal data. The system state information is held in volatile memory.

If the power to a node fails and a backup power source is not available, the system state information is lost which can result in data loss. To avoid loss of data, ensure that each node is powered by an uninterrupted power source (UPS). If mainline power fails, the UPS provides power and time for you to issue the svctask stopsystem command, answer yes to accept the confirmation message, and start the system shutdown process. As the system shuts down, the node loses access to volumes; however, data is not lost. While the UPS supplies power, the node writes cache data and the configuration information that is held in memory to its boot drive. The UPS for the node must be able to provide enough power to allow the shutdown process to complete.
  • If the boot drive is an HDD drive, allow at least 3 minutes.
  • If the boot drive is an SSD drive, allow at least 2 minutes and 30 seconds.
Attention: System data might be lost if you do not issue the svctask stopsystem command and accept the confirmation message to shut down the system (or create an external system to issue the command automatically). Data might also be lost if the node does not have enough power to complete the shut-down process.

Figure 1 shows an example of a system that contains four nodes. Node 1 is elected as the configuration node. The configuration node can be thought of as the node canister that controls the updating of system state. If the configuration node fails, the system can elect a new node to take over its responsibilities. When a change is made to the configuration or internal system data, the same change is applied to all node canisters. This state-driven model of clustering ensures that all nodes in the system know and maintain the exact system state at any time.

  1. A user request (1) is made that results in a change to the system configuration.
  2. The configuration node, Node 1, controls the updates to the system state (2).
  3. The configuration node then forwards that state change to all node canisters, including Node 1. All nodes then make the state change at the same time (3).
Figure 1. Relationship of a system, nodes, and the system state
This figure shows a system with a configuration node, nodes, and the system state.