Application Node Setup: Nodes That Use External Power Controllers
Intended for use with Cassatt Active Response V5.0
Application nodes are servers that Cassatt Active Response deploys into tiers to run your business applications. When used with external power controllers, the application node itself must possess configurable power settings in the BIOS (or exhibit appropriate power behaviors) in order for Cassatt Active Response to actively power manage it. This article describes how to qualify, set up, and configure application nodes (SPARC or Intel) for use with external power controllers. Tasks include:
- Qualify application nodes for use with external power controllers
- Cable hardware
- Configure application nodes
- Add application nodes to external power controllers
Prerequisites
Before starting tasks in this document, review the requirements for completing the tasks:
Qualify application nodes for use with external power controllers
To use Cassatt Active Response Active Power Management, application nodes that are plugged into external power controllers must be able to power on in response to an external stimulus. This is accomplished in Cassatt Active Response through power behaviors on the application node such as "Wake on LAN" and "node boots when power is restored." You also need to assess your goals for active power management in Cassatt Active Response. For example, if you are using Cassatt Active Response in a development/test environment, you may have older and more diverse equipment that may not have one of the recommended power behaviors (or the behaviors are simply non-functional). In a development/test environment, this might be acceptable if you can tolerate occasionally powering on nodes when they do not boot. On the other hand, if you are implementing Cassatt Active Response in a production environment, you may have more stringent requirements and goals for node power management. In either case, you must validate the power functionality on each application node you're going to use with external power controllers—so you understand how the node behaves under Cassatt Active Response power management.
Recommended application node power behaviors/settings
This following table describes node power behaviors/settings that work with Cassatt Active Response active power management, which ones do not, and important considerations.
Cassatt has observed that 1) not all nodes have BIOS settings for power behaviors—sometimes the behavior is just how the node works and is not configurable 2) some settings just don't work, even after you've configured them properly. That's why it's critical to validate application node power behaviors before deciding if a node can be used with external power controllers.
Node behavior or BIOS setting |
Description |
Recommended? |
Suited For...
|
| Wake on LAN |
Wake on LAN is an Ethernet networking computer standard that allows a machine that has been shut down to be booted remotely.
Also called: Remote Wake Up |
Yes |
Production and non-production environments |
Node boots when power is restored
|
If a node has this behavior, when power is taken away from the node, and then power is restored, the node boots, regardless of its previous state. Unfortunately, the name for this behavior/setting is not consistent across application nodes; similarly, although you may find this setting under Power Management in the BIOS, it's not always the case.
Also called: AC Power Recovery |
Yes |
Production and non-production environments |
Node returns to last power state |
This is the default behavior for many nodes. In this case, the node remembers its previous state. If the node is off when the power is removed, when the power is restored, the node remains off and won't boot. However, if the node is on at the time the power is removed, and then power is restored, the node boots. This often appears as random power behavior on the node if you are not aware that this function exists.
This power behavior may be acceptable, but both of the following are required to make it work:
- You must be willing to tolerate inconsistent node power states, and be willing to manually manage those nodes when they don't automatically power on. (This is not recommended in a production environment because it depends on the node always remaining on in the event power is cut at the external power controller outlet.)
- The OS on the node must not have the ability to power off the node after it is shut down; if the OS powers down the node after shutdown, there is nothing Cassatt Active Response can do to monitor and counteract the OS behavior.
|
Yes,
but with caveats |
Non-production environments only
|
| Node does not boot when power is restored |
This is another common default behavior for many nodes. In this case, if the power is taken away from the node, and then power is restored, the node will not boot, regardless of its previous state. Cassatt Active Response will never be able to power on the node without manual intervention. Although not recommended, there may be particular nodes for which manual power intervention is acceptable. You just need to make sure the behavior is known to whoever is managing the node so time is not wasted on troubleshooting. |
No |
Non-production environments only |
Follow these steps to validate the power behavior/settings for all application nodes that you wish to use with external power controllers; this avoids unpredictable behavior during operation.
- If available, enable the Wake on LAN or "Boot when power is restored" setting.
- Test that the setting is functioning using the table below.
- If the setting is not functional, or there isn't a setting, use the table below to determine the node's power behavior.
Node behavior or BIOS Setting |
How to validate behavior or BIOS setting... |
| Wake on LAN (WOL) or "Remote Wake Up" |
- Power off the node.
- From the control node, execute the following command:
ccwol -m <mac_of_the_node> -b <broadcast_address_of_subnet_NIC_is_on>
NOTE: If the control node on which you are running the ccwol command is on the same subnet as the NIC you are trying to wake up, skip the -b parameter; ccwol will use the local 255.255.255.255 broadcast address.
The ccwol command sends a WOL packet at one-second intervals for 30 seconds. If the node is properly configured and the setting is functional, the node will boot.
|
Node boots when power is restored or "AC Power Recovery"
|
- Power off the node.
- Unplug the node.
- Plug the node into an outlet with power.
If properly configured and the setting is functional, the node will boot.
|
| Node returns to last power state |
- Power off the node.
- Unplug the node.
- Plug the node into an outlet with power.
The node should remain off.
- Power on the node.
- Unplug the node.
- Plug the node into an outlet with power.
The node will boot.
|
| Node does not boot when power is restored |
- Power on the node.
- Unplug the node.
- Plug the node into an outlet with power.
The node will not boot.
|
Step 2: Cable hardware
Use the examples in the next diagram and table as guidelines for cabling nodes to external power controllers and network switches. Although the diagram shows multiple switches for illustration purposes, you can configure application nodes with a single switch.
In the example graphic... |
Description |
Application node A |
This node's power connection is to an outlet on the Baytech100 external power controller, which has redundant network access to switches 1 and 2 (which are managed by Cassatt Active Response). |
Application node B |
This node's power connection is to an outlet on the APC2000 external power controller with non-redundant network access to switch 1 only (which is managed by Cassatt Active Response). |
Baytech100 |
This external power controller has no direct access to the Cassatt Active Response network, but has access to Cassatt Active Response via some "Other" network; therefore, network routing must be set up so the "Other" network can route IP traffic to Cassatt Active Response networks. |
| APC2000 |
This external power controller has direct access to switches managed by Cassatt Active Response. The static IP address for the APC2000 (10.10.170.125) is in the range for the Power Subnet (blue); therefore, the ethernet cable from the APC2000 must be connected to a port on the switch that maps to the Power Subnet; if the ethernet cable is connected to a port in switch for the Cassatt Active Response Subnet (red), the APC2000 will not be discovered. |
Follow these steps to cable the hardware:
- Connect application nodes to external power controllers and record the associated outlet number and MAC address.
- Connect application nodes to the appropriate network switch.
If the first NIC in the application node is not going to be managed by Cassatt Active Response, make sure it is unplugged or plugged into its appropriate network.
top
Step 3: Configure application nodes
Using the Intel or SPARC instructions in this section, configure application nodes to boot from the network and enable DHCP.
Intel
Follow these steps to configure boot order and enable PXE booting (which enables DHCP) for Intel-based application nodes. Steps and settings will vary according to the BIOS and network configuration utilities on your application node hardware.
- Verify that power is ON to the external power controller outlet.
- Connect a console, keyboard, and mouse to the application node.
- Power on the application node.
- Access the BIOS setup program and configure as follows.
| For this BIOS setting... |
Set as follows... |
Ethernet interfaces |
Enabled. |
PXE boot |
Enabled. |
Wake on LAN requests |
Enabled. |
Boot when power is applied or restored |
Enabled. |
Boot order |
Network (set to first).
WARNING: The “Network” option must come before the “Hard drive” option or the node will not boot. If you place “Network” first in the boot order, you gain a small optimization in speed for node discovery and in moving nodes to the free pool. If you are operating in a test environment or you expect to need frequent administrative access to the application nodes, you can set the order to Floppy, CD-ROM, Network. |
Other |
Set other BIOS settings as required by your site, making sure they do not conflict with Cassatt Active Response settings. |
Follow these steps to configure SPARC application nodes.
- Note the following before starting:
RARP bootparams cannot be configured for SPARC application nodes; DHCP must be configured for all nodes that are connected to the Cassatt Active Response network.
The devalias command is used to verify that aliases are working on systems; nvalias is used for creating new aliases on systems. Do not use devalias to create aliases for SPARC application nodes under Cassatt Active Response power management; if you do, aliases are lost after a system reset or power cycle. Create permanent aliases using the nvalias command. For more information, see Chapter 3, Setting Configuration Variables, of the OpenBoot 3.x Command Reference Manual.
- Verify that power is ON to the external power controller outlet.
- Connect a console, keyboard, and mouse to the application node.
- Power on the application node.
- Access the application node via the serial console.
To display all supported OpenBoot parameters, use the printenv command.
- Access the OpenBoot “ok” prompt.
For conserver terminal emulator program: Control-e cl1 (c, el, one)
For a Sun keyboard: Stop-A
- Use the devalias command to view devices: ok devalias
net3 /pci@1d,700000/network@2,1
net2 /pci@1d,700000/network@2
net1 /pci@1f,700000/network@2,1
net /pci@1f,700000/network@2
- Set the following OpenBoot parameters.
For this OpenBoot setting... |
Set as follows... |
boot-device to use DHCP |
ok setenv boot-device net1:dhcp
boot-device = net1:dhcp |
boot-command |
ok setenv boot-command boot -r
boot-command = boot -r |
auto-boot? |
ok setenv auto-boot? true
auto-boot? = true |
auto-boot-on-error
(if available) |
ok setenv auto-boot-on-error? true
auto-boot? = true |
local-mac-address
(if available) |
ok setenv local-mac-address? true
local-mac-address? = true |
network-boot-arguments
(if available) |
ok setenv network-boot-arguments dhcp
network-boot-arguments = dhcp |
- Configure at least one the following settings; preferably both.
For this OpenBoot setting... |
Set as follows... |
Boot when power is applied or restored |
Enabled.
NOTE: If the application node has a physical power on/off switch that can be toggled, make sure it is in the On position. |
Wake on LAN requests |
Enabled |
- If application nodes are Sun Opteron servers, disable the High Precision Event Timer (HPET) or application nodes will fail discovery, inventory, and diagnostics. For instructions, see the Sun documentation.
top
Step 4: Add application nodes to external power controllers
Follow these steps to add application nodes to an associated external power controller for discovery by Cassatt Active Response.
Prerequisites:
- External power controller name as defined in Cassatt Active Response
- Outlet number on the external power controller
- MAC address of application nodes
- Go to: Discovered Pool > Power Controllers.
- Click the appropriate external power controller link.
- Select Associated Nodes > Add Nodes and complete the page.
- On the Nodes list, verify that nodes are discovered.
During booting, some Sun systems with gigabit cards may display the following alert. This is normal; booting will eventually start.
Boot device: /pci@1f,700000/network@2:dhcp File and args: -r 1000 Mbps FDX Link up Timed out waiting for BOOTP/DHCP reply Timed out waiting for BOOT/DHCP reply Timed out waiting for BOOT/DHCP reply 1000 Mbps FDX Link up.
Managing nodes that use external power controllers
The following table describes how to manage external power controllers and their nodes after they are up and running.
| Task |
Steps |
Add (initial, manually enter) |
Prerequisites: A discovered external power controller, MAC addresses, and outlet numbers.
- Go to: Discovered Pool > Power Controllers.
- Click the appropriate external power controller link from the list.
- Select Nodes > Associated Nodes > Add Nodes and complete the form.
|
| Add (initial, using a configuration file) |
Prerequisites: A discovered external power controller, MAC addresses, and outlet numbers.
- Create a comma-separated file using the format outlined in the sidebar Configuration files for power controllers and nodes.
- Go to: Discovered Pool > Power Controllers.
- Select the appropriate external power controller from the list.
- From the Power Actions menu, select Add Nodes.
- Click Provide via file, enter the path to the configuration file, and complete the wizard.
|
Update properties |
Updating existing node information such as outlet number, rack, slot, and description, must be done in the Controller by editing the node properties; you cannot use a configuration file to update existing node information. |
top
The following table provides help in troubleshooting discovery, inventory, and managing external power controllers and associated nodes.
State |
Problem |
Follow these steps... |
Inventory |
Nodes are not inventorying |
If nodes fail to inventory:
- Clear any alerts and restart the inventory process and monitor the node.
- If the node does not boot, check that BIOS is configured correctly (at least one of the following must be set, preferably both: boot on power and restore, and Wake on LAN).
- If the node boots, check that the IP address is correct.
- Check that the node is plugged into the right outlet on the external power controller.
- Look at both the node power state and the external power controller outlet power state. Does the outlet power on?Does a different outlet power on? If the outlet powers on, but the node does not, there's a node configuration problem.
|
Up and running |
Alerts for several nodes on an external power controller |
If you get alerts on several nodes attached to an external power controller, there is a problem with 1) the external power controller itself or 2) the configuration of the external power controller.
- Validate the configuration of the external power controller including type, IP address, network, and nodes.
- If the configuration is correct, log into the external power controller and check that the outlets are working.
- Try restarting power on the external power controller. If this doesn't work, follow instructions to replace the failed external power controller.
|
top
Was this article useful? Tell us what you think.
Email infocentral@cassatt.com.
|
