BFC: Correcting Issues Booting from pxe

Reference Information › CA AppLogic Support Knowledge Base › Overview of Support Knowledge Base › Backbone Fabric Controller Home › BFC: Correcting Issues Booting from pxe

BFC: Correcting Issues Booting from pxe

This document covers the procedure for correcting problems occurring if the pxe configuration files under /tftpboot/pxelinux.cfg in the BFC get lost or damage

Problem Description

Any node in a 3.X grid follows these steps when booting:

DHCP Discover is sent to the backbone network
The BFC responds and the node acquires its IP address
The node then proceeds to determine how to boot based upon its boot configuration. This one is stored under /tftpboot/pxelinux.cfg at the BFC in a file for each node, as well as a file named default which indicates the boot method if no specific configuration file exists for the node attempting to boot

If there is no default file and if there is no file corresponding to the node trying to boot, the boot process will fail with messages similar to the following:

Trying to load :pxelinux.cfg/C
Trying to  load:pxelinux.cfg/default
Could not find kernel image:linux
boot:

And the node will simply hang, unable to determine how to boot

Proposed Solution

If no files are left in the /tftpboot/pxelinux.cfg directory, you should create the following ones:

For each one of the servers already present in the grid, create a file whose name corresponds to the hexadecimal representation of its backbone IP address. For instance, a server whose backbone address were 192.168.1.120 would require a file called C0A80178, since 192=C0, 168=A8, 01=01 and 120=78 in hexadecimal. You can use the following link to calculate that representation easily: http://www.hpcfactor.com/support/toolkit/hextools/ip2hex.asp On booting from pxe this is the first file each node will use to boot, if present. Supposing there is already an AppLogic image installed and the node should boot from local disk, its contents should be the following:
```
PROMPT 1
TIMEOUT 10
DEFAULT localboot
LABEL localboot
   LOCALBOOT 0
```
If the server can't find a file with a filename that matches its IP representation, it will try to use the one for the backbone, if present. This pxe configuration file must have a name which matches the network part of the backbone addresses in hexadecimal representation. In the above example, the backbone is 192.168.1.0/24, so the filename would be C0A801. Since this will be used to boot the servers being added to the grid, its contents will most likely point servers to boot from the utility image:
```
PROMPT 1
TIMEOUT 10
DEFAULT utility
LABEL utility
  KERNEL utility-kernel
  APPEND initrd=utility.gz controller=192.168.1.252
  IPAPPEND 3
```
The controller address must be set equal to the BFC address in the backbone. This file may be automatically regenerated by switching off and on manual discovery in the Administration tab of the BFC. The KERNEL file name specified in the KERNEL entry of the file corresponds to an existing kernel file or symbolic link pointing to one in the /tftpboot directory (for instance utility-kernel may be a link: utility-kernel -> vmlinuz-3.2.2-22.xen0 in /tftpboot) .
It is finally advisable to create a default file for pxeboot of the servers which do not match any of the previous criteria. In general those will be directed to boot from the utility image and so their content will be akin to that of the previous file:
```
DEFAULT utility
PROMPT 1
TIMEOUT 10
LABEL utility
  KERNEL vmlinuz-3.0.4-2.xen0
  APPEND initrd=utility.gz controller=192.168.1.252
  IPAPPEND 3
LABEL localboot
  LOCALBOOT 0
```
These files need to be owned by bfcadmin:bfc with permissions 664 for the first two, and by bfcutil:bfc with permissions 644 for the default one.