[hobbit] Looking for suggestions on hobbit SLA reporting

[Henrik Stoerner]

On Wed, Aug 30, 2006 at 04:10:41PM -0400, Kauffman, Tom wrote:

> First -- from a management perspective, our primary application (SAP
> R/3) is either available for use or down. We have a stated objective of
> some % uptime. Scheduled outages are charged just like unscheduled. So
> if we take the system for a scheduled 12-hour upgrade, we get charged
> with 12 hours against our total availability.
> 
> This means that 'blue' status is the same as 'red' -- the guys in the
> field can't use it, it's not available, and management doesn't care that
> it's a scheduled outage.
[snip]
> Henrik, is there a (relatively) easy way to convince the sla report
> funtion to consider 'blue' and/or 'clear' as 'red' for reporting?

You'll have to change the Hobbit source to do that, but it is fairly
simple. In the lib/availability.c file around line 500 is where it
decides what colors go into the availability percentage. That line
reads:

  repinfo->reportavailability = 
     100.0 - repinfo->reportpct[COL_RED] - repinfo->reportpct[COL_CLEAR];

As you can see, "clear" and "red" go together. So you just need to add
"blue" to the list, so it becomes

  repinfo->reportavailability = 
     100.0 - repinfo->reportpct[COL_RED] - repinfo->reportpct[COL_CLEAR] - repinfo->reportpct[COL_BLUE];

> The second problem -- 'red' doesn't always mean "it's dead, Jim"; it can
> also mean "somebody better look at this quick, BEFORE it dies". But we
> don't want to page on yellow, even on this system.
> 
> Suggestions?

Well, that *is* what the yellow color was intended for. You do know that
you can configure the alerts to go out on yellow for just a single
system or test?

> (Is there any way to add 'hot pink' as a status between 'yellow' and
> 'red'? :-)

Or "orange" ... in theory, yes - but I am not at all sure how many
places in the Hobbit code would need changing to make sure they are
all right when there's suddenly one more color to handle.


Regards,
Henrik