IBM Power Systems 775 Manual

Contents

300 IBM Power Systems 775 for AIX and Linux HPC Solution5.4.5 A+ definitionsAlthough product names changed from FIP to A+, the definitions and functions of theproducts remain the same, as shown in Table 5-1.Table 5-1 A+ definitionsThe administrator must set up xCAT A+ node groups that must work with the A+ environment.One xCAT node group called “Aplus_defective” must be set up for any found A+ defectivenodes or octants. A second xCAT node group “Aplus_available” must list the A+ availablenodes or octants. You use the xCAT mkdef command to create the node groups, and then usethe chdef command to associate any node (octants) to the proper node group.The following commands are used to define an A+ group and add a failed resource: mkdef -t group -o Aplus_defectiveCreates an Aplus_defective group that must be empty. mkdef -t group -o Aplus_available members="node1,node2,node3"Create an Aplus_available group with node1, node2, and node3. chdef -t group -o Aplus_defective members=[node]Adds a failed A+ node to the Aplus_defective resources group.5.4.6 A+ components and recovery proceduresThis section describes the tasks that are performed by the administrator or cluster user togather problem data or recover from failures.Definition DescriptionA+ / Fail in Place Component All A+ features including Octants and fiber optic interfaces.A+ / Fail in Place Event A failure event that involves an A+ component or FRU elementthat is left in the failed state in the system.A+ / FIP Refresh Threshold The minimum number of A+ components is available and at thatpoint a hardware replacement is required. The threshold isdetermined from a table in which the values are set according tothe contract policy, expected failure rates, and the amount or timethat is remaining on the maintenance contract.There are individual thresholds for different failure typesA+ / FIP Reset Threshold The minimum number of A+ components that are needed torestore the system to following the repair of A+ components. Theamount of hardware that is replaced is determined from a table inwhich the values are dependent on the component, and theamount of time that remaining on the service contract.There are individual thresholds for different failure typesCompute QCM/Octant/Node A QCM/Octant/Node without I/O adapters assigned to it. It isused solely for running application code, and often runsdegraded.Non-compute QCM/Octant/Node This is a QCM/Octant/Node with I/O adapters assigned to it. It isused for disk or I/O access and often must retain full function.