Infrastructure Director

This section describes known issues of the N1 Provisioning Server Infrastructure Director.

Failures During Disk Copy Operations When Farm Goes Into Standby

When a farm goes into standby mode, disk copy operations for all servers belonging to that farm start simultaneously. However, when a failure occurs during the first disk copy operation, for example, not enough disk space, the other disk copy operations continue until all are completed, and then the failure is reported. To resolve this, ping the farm using the command farm -pf farm-ID. Then put the farm in standby using the command farm -s farm-ID. (4849694)

No More Resource Exception When No Servers Available in IDB

When no more provisionable servers are available in the IDB, you will not be able to move a server from one server group to another in a single farm operation. Remove the server from the first server group and update the farm. Then add the server to the second server group and update the farm.

The server might be provisioned by another farm between removing it from the first server group and adding it to the second. Consequently a “no more resources” exception will occur.

(4849699)

shelfsync Categorizes by Type Not Size

Even though the discovery mechanism detects differences in the blades, such as memory size and number of CPUs, these differences are no currently a part of the selection criteria for CPU resources for a farm.

If you need this information, use the device and devtype commands to create pools of like-resources from the discovered items that have the common elements that the original resources use. (4849705)

Rebooting Provisionable Servers Overwrites DNS Setup

The N1 Provisioning Server software overwrites the DNS setup information in the /etc/resolv.conf configuration file when the provisionable server is rebooted. Any changes to that file, will be lost when rebooting the server. Create a boot time startup script that will restore the DNS setup information that you specified previously. (4849715)

mls Command Does not Show Accurate Status

An mls -a command might inaccurately report the agent on a server as being marked DOWN. Wait 60 seconds and try again to confirm the status of the node. The normal monitoring of the node by the control plane server is not affected by this condition, and monitoring will accurately report a failed node. (4849719)

Command Usage References to Firewalls and Load Balancers

Some commands might have references to firewalls and load balancers in their usages as displayed in the man pages. Any command options and usage references to firewalls and load balancers do not apply to the Blades Edition of the N1 Provisioning Server software. Ignore these references. (4856855)

Adding a Backhaul Device to a Farm

When adding a backhaul device to a farm, make sure that the port on the switch of the shelf that is reserved for the backhaul device is not in trunking mode. Typically, after bench configuration, the ports on the shelf are in trunk mode. Change this when adding a backhaul device to these ports. (4856867)

Deleting Active Requests Before Stopping the IDB

Make sure that there are no active requests in the system before stopping the IDB server. If the IDB server fails, existing Farm Manager processes sometimes do not exit gracefully. Before restarting the IDB server, stop all remaining Farm Manager processes by performing the following steps:

1. Check the existence of any Farm Manager processes by executing the following command:

   /usr/ucb/ps -auxwww | grep -i “com.terraspring.cs.fm”

2. Use the UNIX kill command to stop any remaining Farm Manager processes.

(4856872)

System Controller Hangs During Power State Change

During operation on a farm with a large number of servers, the system controller sometimes hangs indefinitely. If this happens, perform the following steps:

1. Stop the Farm Manager process for the affected farm. To determine the process identification (PID) of the corresponding Farm Manager process, type /usr/ucb/ps -auxwww | grep "com.terraspring.cs.fm". In the list that displays, look for the process with the farm ID in the command-line arguments.

2. Use the kill command (or kill -9 if necessary) to stop the identified process and any child processes.

3. Power cycle the hung chassis. The system controller should be accessible after the power cycle. If not, contact your technical support representative.

4. All server blades on the power-cycled chassis will lose power and replaceFailedDevice requests might be filed before the blades are done rebooting.

5. Restart the aborted farm operation.

6. Delete any replaceFailedDevice requests on the rebooted chassis by executing request -d request-ID.

(4856873)

Power Command Issues

The power command with the -off option is similar to the UNIX command poweroff, which powers off the device on which it is issued. When using the power command with the -off option, be sure to have a blank space between power and -off, otherwise you will power off the control plane server. (4857749)

Securing DNS Zones for Farms

To make DNS zones for every farm secure from each other, set the com.terraspring.dns.useacl property to true in the /etc/opt/terraspring/tspr.properties file. This property is set to false by default. After changing this property, issue the /opt/terraspring/sbin/dnsmod sync command to restart the DNS server. (4857751)

Configuring Port Connectivity When Using a Separate Image Server

For an I-Fabric that has fewer than three chassis switches and a separate image server, connect the image subnet interface on the image server to the NETP1 port on the first chassis switch. Then, configure the NETP1 port of that chassis to the image VLAN (vlan 8).

Issue the following set of commands to configure the NETP1 interface of the Sun Fire^TM B1600 chassis switch:

Vty-0#configure
Vty-0(config)#interface ethernet NETP1
Vty-0(config-if)#description External RJ-45 connector NETP1 
Vty-0(config-if)#switchport allowed vlan add 8 untagged
Vty-0(config-if)#switchport native vlan 8              
Vty-0(config-if)#switchport forbidden vlan add 2
Vty-0(config-if)#end

If there are two chassis switches, connect NETP0 of the second chassis switch to NETP2 of the first chassis switch. Both of these ports will be configured automatically during the installation process.

For an I-Fabric that has more than three chassis switches and a separate image server, connect the image subnet interface on the image server to a port on the external data layer switch. Then, configure that port of the external data layer switch to the image VLAN (vlan 8) and add the image VLAN to the VLAN database.

The image VLAN configured on this switch should be vlan 8. Type the following to create vlan 8:

enable
vlan database
 vlan 8 name ImageVlan
 exit

For example, on the Cisco 3550 and 4503 switches, you would issue the following set of commands to achieve this connection for port GigabitEthernet 2/19:

qe1-4500>enable
Password: 
qe1-4500#
qe1-4500#conf terminal
Enter configuration commands, one per line.  End with CNTL/Z.
qe1-4500 config)#interface GigabitEthernet 2/19
qe1-4500(config-if)#switchport access vlan 8
qe1-4500(config-if)#^Z
qe1-4500(config-if)#write mem

(4857752)

Updating Images Requires New Image Name

When you update an existing image, a copy of the original image is placed onto a new disk to allow you to upgrade the newly copied image. Before releasing the new image using the image -r command, you must give the new image a unique name. (4877335)

Replacing Failed Blades

Before replacing a failed blade with another blade, ensure that you have free blades available by executing the device -l command. (4919188)

Adding a New Blade to an Existing Chassis

When adding a new blade to an existing chassis, the new blade is not configured by the shelfsync command. Follow the steps described in Chapter 3 of the N1 Provisioning Server 3.0, Blades Edition, System Administration Guide to configure the newly added blade. (4919189)

Changing Power State of Provisionable Servers Used in an Existing Farm

Do not independently power on or off provisionable servers in an existing farm. (4919199)

Listing Provisionable Servers in Use

Sometimes the command device -L -U does not list all used servers in the system. A better way to determine all used servers in a system is to issue the command device -l | grep -i used. (4919200)

Specifying a JumpStart File in the image Command

Ensure that the JumpStart^TM file exists before you specify it as a parameter of the -j option in the image command. Otherwise, a java.lang.NullPointerException is printed out with a stack trace. (4919203)

tsprmonitor Usage

Although the usage information of the tsprmonitor command indicates that the time parameter is optional, this parameter is mandatory. Therefore the usage should read as follows:

tsprmonitor -stop minutes-to-be-rebooted [-c]

(4919204)

Gigabit Ethernet Card Instance Assignment

The gigabit Ethernet card must be assigned an instance of 0. (4924060)