[Orca-users] Orca and AIX

[David Michaels]

We do have an AIX-allator, thanks to Rajesh and some mods from myself 
and others.  However, it doesn't do anything at the frame level.

I think what you describe might be doable with the existing setup.  I 
think you can just modify your orcallator.cfg file on your orca server.  
You'd probably need a special group, so that it would put data from 
multiple hosts into one graph.  Specify a line_type of "stack".

For example, instead of:
find_files     
/orca/path/var/orca/orcallator/*(.*)*/(?:(?:orcallator)|(?:percol))-\d{.......
                                              ^^^^
try:
find_files     
/orca/path/var/orca/orcallator/*.**/(?:(?:orcallator)|(?:percol))-\d{.......
                                              ^^

I'm really not sure how well that would work, though.  If it doesn't 
work, you may have to run a separate script to consolidate the CPU usage 
columns into a separate data file, with a copy of the column for each 
host.  Then you can graph those columns stacked.

Alternatively, modify the aixallator to tag the CPU data column with the 
hostname.  Then set your data field in the corresponding plot {} to call 
it as such.   e.g.:

plot {
title      Frame CPU usage
data     ((?:[^_]*))_cpu
line_type    area
line_type    stack
legend    $1
y_legend    Cumulative CPU
data_min    0
data_max    100 * 64
plot_min    0
plot_max    100 * 64
}

Blair or someone else on the list might know of a slicker way to do this 
with the orcallator.cfg than I've described.  These ideas are just my 
best guesses without having the chance to test it out.

Cheers,
--Dragon

Angelo McComis wrote:
> Hello,
>
> I'm new on the list, so if this is already covered somewhere, please 
> feel free to redirect me to the appropriate links instead of typing a 
> lengthy reply.
>
> I'm looking to use Orca within my pSeries environment to get per-frame 
> stats rather than at the individual LPAR level.
>
> The problem I'm looking to solve is that, for example: Let's say I 
> have a pSeries frame with 64 total CPUs. I have 10 LPARs with 6 CPUs 
> but all 10 can have up to 8 CPUs. Obviously all 10 cannot have 8 CPUs 
> each at the exact same time.  I would be looking for a graph (let's 
> say scale of 1-100) with a solid line at my max CPU available in the 
> frame (e.g. 64), and a stacked/filled area graph representing the 
> usage of the 10 LPARs within that space. If/When the graph flatlines 
> at 64, I know I'm overcommitted, and know which LPAR needs to move to 
> have more room.
>
> Any direction at all on this is greatly appreciated.
>
> Angelo
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Orca-users mailing list
> Orca-users at orcaware.com
> http://www.orcaware.com/mailman/listinfo/orca-users
>   

-------------- next part --------------
An HTML attachment was scrubbed...
URL: </pipermail/orca-users/attachments/20080908/a00bb260/attachment.html>