Was wanting to solicit feedback from people who have been doing things with 
confluent.  In terms of general 'look and feel', where xCAT core and confluent 
are comparable, is confluent going in a good direction, or is it missing the 
boat?

Broadly speaking, we are trying to decide how to prioritize further changes, or 
if there are changes people want that we are not doing.  Some of the concepts:

-nodestorage command - Likely going to be going in the 2.2 and a fairly 
straightforward thing to provide, a command to manage the hardware raid, 
remotely at least on Lenovo equipment, to create/remove arrays and volumes, 
designate disks as hotspares, that sort of thing

-discovery bridge to other software - Provide a way for confluent discovery to 
feed into external OS deployment software.
-xCAT: For example, nodeadd with mac address and other attributes and request 
the 'chain.chain' to run on xCAT terms.  This would eliminate the need for 
dynamic dhcp ranges, as confluent's PXE discovery doesn't need addresses to work
-Other deployment engines

-Integrated os deployment capability  This would mainly sidestep the hassle of 
having two distinct yet similar software being configured. However a few 
differences currently in my mind:
                -Dispense with database-hosted metadata about images and 
organize everything in filesystem instead, using symlinks and such to avoid 
duplicating data rather than database records pulling them together
                -Designed to enable more security in deployment (https with 
validation, support non-PXE install) with easy and a few tunables to adjust to 
opt into 'riskier' but more universal approaches.
                -Have a bit more fleshed out builtin network configuration 
facility
                -Move SSH strategy away from standalone keys to ssh ca keys 
(supported by all reasonably recent linux distros by now)
                -Easy tunable to opt into a popular configuration management 
engine of users choice (probably ansible and salt to start)
                -Move away from having our own open-ended customizable 
'postscripts' facility to rely instead on the above for open ended postscript
              -Stateless would probably be a port of xCAT's existing genimage 
facilities, I don't think I have a lot of ideas to make that any less complex 
than it is personally...
                -No longer loop-mount .iso files
                -Potentially using a somewhat different template syntax, to 
enable use of a more universally used templating engine.


-More web development.  Currently the web ui is mainly a reduced function 
dashboard, relying heavily upon CLI access for real operation.  If people want 
a richer web gui, we may need to prioritize.  Otherwise it's pretty much going 
to be about on the level it already is.
_______________________________________________
xCAT-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/xcat-user

Reply via email to