I am thinking that a few flags will be standardized across commands. I've
already done nodesensors with the following:
-c: Lay things out in a reasonable CSV format. Similar to piping to xcsv
(Why not just pipe to xcsv? Two things, one is that nodesensors will move
units data to header instead of next to value, allowing the output to be
strictly numeric for easier import into a spreadsheet.
For another, time series data is supported, and it's easier to automatically
adapt to that within the application itself).
I'm considering another:
-u: group output by unique value . Similar to pipe to xcoll. In this case the
main benefit would be that in trying to find nodegroups, it would be nice to
use the incoming noderange for
hints to guide tie-breaking of equivalent appropriate groups. Other than
that I would probably just make a generic analyzer. I'm torn here because I
see a lot of flags that would be
useful that would make me lean toward an enhanced xcoll. Namely:
As you go: For every new bit of output, clear screeen and redo
against current state of output
No-gather: If you just wanted to organize the output in an
'xcoll' way, but not identify sameness
Diff output: Identify the 'modal' output and then present
non-modal output with diff style indication of what makes those different
Group count: xcoll -c like, count of nodes that gave that result.
No abbreviation: (-n in xcoll, making it utterly standalone with
no dependency on xCAT or confluent)
Occasionally the abbreviation of the incoming noderange is odd and could
look less odd if it knew the incoming noderange, but I don't know about how
important that is versus the clutter of trying to accommodate all the behaviors
above in 'universal' flags on various commands.
------------------------------------------------------------------------------
_______________________________________________
xCAT-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/xcat-user
