Configuring confluent
When used in conjunction with xCAT
If you wish to use xCAT to create the confluent configuration, see Configuring confluent from xCAT The remainder of this document is mostly applicable to users working with confluent standalone directly rather than using xCAT to configure.
Using Confluent without xCAT
In confluent, configuration is organized as attributes on nodes. The attributes may be directly configured on a node or inherited from a group. Values may be a straightforward string or an expression as documented in the attribute expressions documentation.
To get started, it is suggested to use the everything
group to set universal attributes.
The everything
group automatically exists and all defined nodes are automatically
placed into it, in addition to manually specified and created groups
Usually a cluster uses the same username/password across the XCCs (xClarity Controllers). In such a case, it is
suggested to set this data as attributes on the everything
group:
nodegroupattrib everything console.method=ipmi
nodegroupattrib everything -p bmcuser bmcpass
The nodeattrib
and nodegroupattrib
commands can either accept key=value
or -p key1 key2
. The former allows fully non-interactive use, while the
latter prompts interactively to prevent the username and password from being inadvertently on screen or stored in your shell history.
From there, adding a specific node using values from the group everything
combined with node specific attributes could involve the following:
nodedefine n3 bmc=n3-xcc
Another common task is to create a custom group, with particular meaning to a specific environment. For example, here is creating a custom group called rack1
and
using the expression syntax to append ‘-xcc’ to the end of the nodename to use as the XCC address:
nodegroupdefine rack1 location.rack=1
nodegroupdefine compute bmc={nodename}-xcc
These groups, like the everything
group can hold any attribute, and may also use expressions or normal values. The process to create a node can include these groups:
nodedefine n1 groups=rack1,compute
At which point, n1
has location and XCC address configured just by virtue of the nodes it was assigned to. Note that membership in
the everything
group is automatic, even if not listed in the groups for a node to be in, it will nevertheless be considered a member of that group.
At this point, there are a few alternative paths to proceed:
- Set up confluent to be able to deploy operating systems (recommended prior to auto-configuration if planning for OS deployment using confluent): Preparing for Operating System Deployment
- Discovery is intended to help when IP addresses, usernames, and/or passwords are not configured yet. If these activities are otherwise handled in the environment, it is suggested to skip discovery and read: Managing hardware using confluent
- If the IP addresses, usernames, and/or passwords need to be configured and it is desired to use physical location of equipment as the key, see: Using switch based discovery for rackmount servers or Using enclosure based discovery for dense platforms
- If the IP addresses, usernames, and/or passwords need to be configured and it is desired to either use serial numbers, mac addresses, or otherwise manually review the available data to proceed, see: Using nodediscover assign