Before The Kernel Loads; Provisioning Nodes; The Kernel Image, Ramdisk And Kernel Modules; Node-Installer; Node States; Updating Running Nodes; Adding New Nodes; Troubleshooting The Node Boot Process
Thanks to @Andrew Hill11 for helping me puzzle through this. The problem was that ConnectX-2 cards, like a lot of Mellanox cards, have trouble booting from http. Use the following sets of commands to switch back to tftp:
# cmsh
% category use <your-category-name>
% set bootloaderprotocol tftp
% commit
This article explains Bright Cluster Manager can be configured to manage a regular node with a Self Encrypted Drive (SED), so that the regular node is able to boot via PXE and be provisioned.
This procedure was tested on... (More)
In recent (december 2016) versions of RHEL/CentOS 7.3, the bridge kernel module has been split into two components: the bridge and br_netfilter kernel modules. By default, Bright adds the bridge kernel module to the list of kernel modules in any... (More)
How do I revert back to PXE boot over TFTP instead of HTTP?
Starting from Bright 7.3, HTTP is used for nodes PXE booting instead of TFTP. However, it has been noticed that for some hardware the nodes hang, and... (More)
