[Xymon] Memory monitoring - capped zone
Vernon Everett
everett.vernon at gmail.com
Tue Jul 24 14:04:15 CEST 2012
Hi Henrik
I was hoping you would respond. :-)
My comprehension of C is shaky at best, which is why I asked the list.
Here's my understanding of the problem, and how I think we can improve
things.
I am happy to code any changes to the script side, but there may be some
updates required in the C code.
Using prtconf to get installed memory makes sense.
In a capped zone, it will show the cap.
In an uncapped zone, it will show the total memory, as seen from the global
zone. (This may not really be the available memory because of other zones,
but that is the potential memory available, and the best we can get at the
moment.)
However, whenever we check the memory within a zone, capped, or uncapped,
it interrogates the kernel. The kernel returns all the free memory it can
see. From the global zone!
The problem arises in cases where there is more memory free than the cap
size.
e.g. 2Gb cap, with 3Gb free memory.
So how to establish how much memory is really free in a capped zone.
If we run prstat -Z, the last lines will show the memory footprint of the
zone.
(If you run it in the global zone, it shows the memory footprint of all the
zones.)
The memory cap, less the footprint, should give us the free memory within
the zone.
I do not believe there is any way from within a zone, to determine if you
are capped or not capped, but that shouldn't make a difference. However,
it's easy to deteremine if you are in a zone.
"zoneadm list" should be a single line, not equal to "global".
A non-zero exit code, or a return of global, indicates you are in a global
zone, or a system not running zones.
Translating to pseudo code
if [ is a zone ] and [ not global zone ]
then
use prtconf to get memory footprint
calculate free memory using prtconf and memory footprint
else
use existing method to determine free memory
fi
Thoughts and feedback appreciated.
Regards
Vernon
On 24 July 2012 15:58, Henrik Størner <henrik at hswn.dk> wrote:
> On 24-07-2012 05:45, Vernon Everett wrote:
>
>> Hi all
>>
>> I think I have eventually figured out how we can monitor memory in a
>> meaningful way in a capped zone.
>> Looking at the "client data" sent to the server, can anybody tell me
>> which data are used to calculate the figures, and percentage used for
>> memory?
>>
>
> It depends on what operating system is monitored - memory monitoring is
> one of the most messy parts of the Xymon client, simply because there is no
> standard way of reporting it from the various operating systems.
>
> Look at the OS-specific code in xymond/client/*.c
>
> E.g. on Solaris, it uses the "prtconf" data to determine the amount of RAM
> installed, the "memory" data to figure out how much is used, and the "swap"
> or "swaplist" data for swap information (depending on what version of the
> OS and client is installed).
>
>
> Regards,
> Henrik
>
> ______________________________**_________________
> Xymon mailing list
> Xymon at xymon.com
> http://lists.xymon.com/**mailman/listinfo/xymon<http://lists.xymon.com/mailman/listinfo/xymon>
>
--
"Accept the challenges so that you can feel the exhilaration of victory"
- General George Patton
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xymon.com/pipermail/xymon/attachments/20120724/9b85e2a0/attachment.html>
More information about the Xymon
mailing list