Enhanced Support Facility 2.6 User's Guide

1.6.1 For SynfinityCluster

When using SynfinityCluster, you need to set the SCF/RCI monitoring timeout in the kernel parameter (/etc/system) according to RCI connecting unit model or the number of partitions.

Notes

• The monitoring timeout might need to be set for some RCI connecting unit without partitions.
• You can calculate the timeout using the largest number of partitions in an RCI connecting unit.
• When the timeout setting is done, reboot a node and manually set the SynfinityCluster parameter (failure detection monitoring time). See "5.3 Alert monitoring interval" of the "SynfinityCluster Installation/Administration Guide".
  • Model with partitions: See "Condition a. "Model 800, 1000, and 2000".
  • Model without partitions: See "Condition b. Cluster system with 4 or more nodes except the above "a".

For GP7000F model 200/200R/400/400A/400R/600/600, and PRIMEPOWER200/400/600

The monitoring timeout setting is not required.

For PRIMEPOWER 250/450

Set 2 seconds for the monitoring timeout.

• Setting up the /etc/system file

Change the /etc/system file on all cluster nodes, as follows:

1. Copy (or backup) /etc/system using /etc/system.org:

     Example: # cp /etc/system /etc/system.org

2. Add the following to /etc/system. As the timeout is set up in *s units, set a value equal to the value calculated above, multiplied by 1000000:

set FJSVscf:scf_rdctrl_sense_wait = (monitoring timeout: *s unit)
For example: /etc/system is specified as follows:
set FJSVscf:scf_rdctrl_sense_wait = 2000000

3. Reboot the system

For GP7000F model 1000/2000, and PRIMEPOWER 800/1000/2000

Set up the monitoring timeout in the /etc/system file as follows:

• Calculating monitoring timeout
  • 2 partitions : 2 seconds
  • 3 or more partitions: 1 second + (0.5 x number of partitions)
    • Example 1) 3 partitions: 2.5 seconds
    • Example 2) 4 partitions: 3.0 seconds

• Setting up the /etc/system file

Change the /etc/system file on all cluster nodes, as follows:

1. Copy (or backup) /etc/system using /etc/system.org:

     Example: # cp /etc/system /etc/system.org

2. Add the following to /etc/system. As the timeout is set up in *s units, set a value equal to the value calculated above, multiplied by 1000000:

set FJSVscf2:scf_rdctrl_sense_wait = (monitoring timeout: *s unit)
For example, /etc/system is specified for 2-partition configuration as follows:
set FJSVscf2:scf_rdctrl_sense_wait = 2000000

3. Reboot the system

For PRIMEPOWER 650/850

Set 2 seconds for the monitoring timeout.

• Setting up the /etc/system file

Change the /etc/system file on all cluster nodes, as follows:

1. Copy (or backup) /etc/system using /etc/system.org:

     Example: # cp /etc/system /etc/system.org

2. Add the following to /etc/system. As the timeout is set up in *s units, set a value equal to the value calculated above, multiplied by 1000000:

set FJSVscf3:scf_rdctrl_sense_wait = (monitoring timeout: *s unit)
For example: /etc/system is specified as follows:
set FJSVscf3:scf_rdctrl_sense_wait = 2000000

3. Reboot the system

For PRIMEPOWER 900/1500/2500/HPC2500

Set up the monitoring timeout in the /etc/system file as follows:

• Calculating monitoring timeout
  • 2 partitions : 2 seconds
  • 3 or more partitions: 1 second + (0.5 x number of partitions)
    • Example 1) 3 partitions: 2.5 seconds
    • Example 2) 4 partitions: 3.0 seconds

• Setting up the /etc/system file

Change the /etc/system file on all cluster nodes, as follows:

1. Copy (or backup) /etc/system using /etc/system.org:

     Example: # cp /etc/system /etc/system.org

2. Add the following to /etc/system. As the timeout is set up in *s units, set a value equal to the value calculated above, multiplied by 1000000:

set FJSVscf3:scf_rdctrl_sense_wait = (monitoring timeout: *s unit)
For example, /etc/system is specified for 2-partition configuration as follows:
set FJSVscf3:scf_rdctrl_sense_wait = 2000000

3. Reboot the system