1.3 Remote Power Management Mechanism
1.3 Remote Power Management Mechanism
Remote power management for SPARC M12 and SPARC M10 is controlled per remote power management group.
Of the host nodes in a group, the host nodes with the remote power management function enabled are targeted for remote power management. The power supply status of a remote power management group is determined depending on the host node status in the group.
Of the host nodes in a group, the host nodes with the remote power management function enabled are targeted for remote power management. The power supply status of a remote power management group is determined depending on the host node status in the group.
- On state
The power is on to one of the host nodes in the remote power management group. - Off state
The power is off to all the host nodes in the remote power management group.
This section describes the mechanism of interlocking when powering on and off, with the following settings assumed.
Setting Item | Host Node #0 | Host Node #1 | Host Node #2 | I/O Node #0 | I/O Node #1 |
---|---|---|---|---|---|
Interlocking setting | Disable | Enable | Enable | Setting disabled | Setting disabled |
Master node | Yes | No | Yes | Setting disabled | Setting disabled |
Mechanism of interlocking when powering on
If any of the host nodes in a remote power management group is powered on, then all of the host nodes, subnodes, and I/O nodes in the group are powered on. They are powered on in the order of host node and I/O node.
Note - You can set the length of time that the host node waits until I/O node devices are accessible. For this setting, use the setpowerupdelay command of the XSCF firmware For details, see "4.2.1 Setting/Checking the Warmup Time" in the Fujitsu SPARC M12 and Fujitsu M10/SPARC M10 System Operation and Administration Guide. If the length of time that the host node waits is not set, the host node may be unsuccessful in accessing an I/O node device and fail to start the system. In addition, replacing an I/O node or changing the settings of an I/O node may change the time taken until the relevant device becomes accessible. This change in time may prevent the host node from accessing the device. When an I/O node has been replaced or the settings of an I/O node have been changed, use the setpowerupdelay command to set the length of time that the host node waits, again. |
Mechanism of interlocking when powering off
After all the host nodes in a remote power management group are powered off, all I/O nodes in the group are powered off.
Remote power management using Wake on LAN
Generally, the target nodes of the remote power management function for SPARC M12 and SPARC M10 are the hosts and I/O devices on which a controller is mounted. The controller allows IPMI communication even while the power of the hosts and I/O devices are turned off.
When all of the following conditions are satisfied, those devices on which such a controller is not mounted can also perform power interlocking using the remote power management function for SPARC M12 and SPARC M10.
When all of the following conditions are satisfied, those devices on which such a controller is not mounted can also perform power interlocking using the remote power management function for SPARC M12 and SPARC M10.
- Wake on LAN is supported.
Power-on is performed with Wake on LAN. - IPMI communication can be performed.
After power-on using Wake on LAN, IPMI communication through the LAN is used for performing power-off and obtaining the power state. - They are connected to the network on the same subnet as XSCF-LAN#0 of the master host node or XSCF-LAN#0 and XSCF-LAN#1 of the master host node.
Note - A host node with the Wake on LAN setting cannot be the master node. |
Note - Wake on LAN cannot be set for SPARC M12 and SPARC M10. Therefore, Wake on LAN cannot be used to power on SPARC M12 and SPARC M10. |
Note - The Wake on LAN setting varies depending on the node. See the respective node manuals. |
Interlocking at the failure recovery time
If a node in a remote power management group cannot communicate when recovered from a failure or other problem, operation is as follows.
- For an I/O node failure
If the power to the remote power management group is on, the master host node issues a power-on instruction. - For a host node failure
Even if the power of a remote power management group is on, the master host node does not issue a power-on instruction.
< Previous Page | Next Page >