Troubleshooting provisioning issues

The following topics provide information and workarounds related to provisioning:

(WinPE) Problem with injecting driver into a WinPE image

To inject a driver into a WinPE image during image creation, do not include spaces in the path to the .infdriver file. This is true for driver paths in the Drivers.txt file and any driver paths you add with the Select Driver Files dialog box. This limitation is imposed because Windows AIK accepts Drivers.txt as a space separated file.

(WinPE) Network configuration window appears during a WinPE boot

If the network configuration window appears during a WinPE boot, it might be because the DHCP server is taking longer than usual to assign IP addresses. The boot might continue normally if you close the network configuration window immediately after it appears. If you are using the DHCP server to assign IP addresses (not the network.ini file) and if the server fails to assign the network details due to slow initialization of network cards on the target machine, edit the BLAssignNetDetails.vbs file and increase the ATTEMPTS and DELAY values to allow for initialization time. For example, you might increase ATTEMPTS from 2 (the default) to 5 and DELAY from 10 (the default) to 30. The BLAssignNetDetails.vbs file resides in the /provisioning/winpe subdirectory that you created by unzipping the <version>-provision-files.zip file (which you obtained from within the main product package that you downloaded from the EPD site).

(HP-UX or AIX) Provision Job canceled

If the Oracle Real Application Clusters (RAC) node goes down during execution of an HP-UX or AIX Provision Job, the job is canceled. If this occurs, you can execute the Provision Job again.

(SUSE 8) Provisioning SUSE 8 SP4 has problems

The procedure for provisioning SUSE 8 SP4 is different than the normal procedure. Check the support website for updated instructions.

(Solaris) JumpStart process fails because it cannot find a rules.ok file

Provisioning through JumpStart does not work properly in a Domain Name Server (DNS) environment. In an environment where DNS is listed first or second in the host line of /etc/nsswitch.conf, the scripts that generate the fields in /etc/hosts, /etc/ethers, and /etc/bootparams fill in the computer's name with FQDN. The JumpStart process fails when it tries to find the rules.okfile. To fix this problem, use the TrueSight Server Automation Console to specify FQDN as the computer name.

(Solaris) Problems with Additional Profile Entries section in a Solaris system package

Provisioning does not parse the Additional Profile Entries section of a Solaris system package to ensure that restricted flash archive keywords are not used in the text area. Restricted keywords should not be used in this field.

Serialization error when saving the system package due to large kickstart file

After customizing the kickstart file in a system package for ESX, ESXi, Red Hat Linux, or Ubuntu Linux, you are not able to save the system package. When you attempt to save the system package, you receive the error "Failed to serialize message."

This issue occurs when the kickstart file increases in size and is larger than the size limit of 64 kilobytes. Reduce the size of the kickstart file by removing some of its contents. For example, remove comment sections. Then try saving the system package again.

(ESX 3.x) Datastore and storage created even when not requested

ESX 3.x provisioning creates a datastore and storage by default even though no instructions were given in the system package to create the datastore and storage.

(ESX 2.5 on XEON) Incorrect boot image used for virtual instances

OS detection for WinPE-based provisioning can send an incorrect boot image for virtual instances. With ESX 3.0, it is possible to define the type of CPU virtualization (either 32- or 64-bit) while defining the instance. With ESX 2.5, this is not an option as all instances are 32-bit. As a result, ESX 2.5 running on XEON-based hardware always detects the instance as the underlying 64-bit CPU when the hardware probe is done. To fix this problem, disable the 64-bit image as a default and only enable the 32-bit image.

(PXE) Provisioning fails when the PXE server is on a different subnet than the target device

In a certain environment, provisioning might fail if the PXE server and the target devices to provision are on different IP subnets. In such an environment, specifying a multicast address during the configuration of the PXE server does not allow proper communication between the PXE server and the target servers for provisioning, and multicast is not set up in the network.

Workaround: During configuration of the PXE server (either through the GUI or through the blasadmin utility or editing the multicast_address option in the pxe.conf file), specify the unicast IP address for the PXE server, rather than a multicast address.

Note

When you boot the PXE client, the PXE server sends various vendor-specific information (also known as option 43) to the PXE client. This information includes the discovery control (suboption 6), which is 0x40 for multicast communication, or 0x80 for broadcast communication. In addition, the multicast or broadcast IP address is specified in suboption 7.

When you apply the current workaround, these suboptions are similar to those for standard multicast communication, but the unicast IP address that you specify is used.

 

(PXE) Provisioning fails when the PXE service is on the same server as the DHCP server or DHCP is not on the same server as PXE

During installation, the installer issues a prompt if the DHCP and PXE services are on the same system. If the option is not selected correctly, or DHCP is added or removed from the PXE server, then there is additional configuration you need to perform:

  • If DHCP and PXE are on the same system:  Set DHCP Option 60 (PXEClient) and ensure the -Dblx.bmlistendhcp option is set to false in the NSH/br/blpxe script or HKLM\Software\BladeLogic\PXE Server\PXE key.
  • If DHCP and PXE are on different systems: Ensure that the DHCP Option 60 (PXEClient) is not set, and set the -Dblx.bmlistendhcp option is set to true in the NSH/br/blpxe script or HKLM\Sofware\BladeLogic\PXE Server\PXE key

For more information please see Configuring-a-DHCP-server-on-Linux or Configuring-a-DHCP-server-on-Windows

(Windows) Saving system package as XML file causes an exception

If you create Windows 2003 system packages without specifying any values, saving the system package as an XML file results in an exception. You can update the empty Windows 2003 system package with required field values and then save it as an XML file.

(Linux Kernel 2.6) PXE server experiences problems

To run the PXE server on Linux Kernel 2.6 (Red Hat/SUSE), you must disable the IPV6 stack on the interface where the PXE server is configured, as follows:

1. Add the following two lines, if they are not already present, to /etc/modprobe.conf:

alias net-pf-10 off
alias ipv6 off

2. Reboot the server.

(AIX) Character length limitations on optional bosinst attributes in AIX system package objects

When you try to create a filesets bosinst attribute with a long list of custom filesets, the System package object fails, and the save, or close and save operations fail. (This action is usually not a problem on native NIM. See Using the NIM bos_inst operation, in the IBM knowledge center.)

There is a length restriction for the OM database schema: When the value of the SYSTEM_PACKAGE_NIM_BOSINSTATTR.VALUE is more than 255 characters, the Invalid Field Error: 8152 message is displayed.

If you must specify a value greater than 255 characters, define a local parameter in the system package or device, and use it in the bos_inst script. This creates the required file with a long string in the datastore, as in the following example: 

  1. Create a local property and assign it a value greater than 255 characters.
  2. Navigate to the system package and select the Optionalbos_inst attribute.
  3. Assign the local property you created in step 1 to the boot_client.

To verify, run a job and check the bosinst shell file on the NIM server, using the script 5.bosinstOp.sh

Fragmentation of TFTP packets over a WAN

In an environment where the TFTP server communicates with target servers across a wide area network (WAN), a slow WAN might result in fragmentation of TFTP packets. In such a case, you can increase maximum TFTP packet size through the tftp.conf file. To increase maximum TFTP packet size, increase the value of the tftp_packet_length setting in the tftp.conf file. The default value for this setting is 528 bytes, and the allowed range of values is 8 to 65535 bytes. 

Provisioning fails: driver bundle contents

The provisioning process fails if the driver bundle being copied includes a directory with a dot in its name.

BMC provisioning removes preexisting entries from NIM database 

If a machine already exists in the NIM database prior to provisioning with the BMC BladeLogic system, the provisioning process removes it from the database regardless of whether the provisioning job execution was successful.

iDRAC device details do not appear on the console

The content editor of the console does not display device details for iDRAC devices. Details such as CPU Count, CPU Family, CPU Speed, and RAM show a value of 0.

 

Tip: For faster searching, add an asterisk to the end of your partial query. Example: cert*