[csw-buildfarm] Fwd: OpenCSW catalog update report (ganglia)

[Daniel Pocock]

Just adding to my comments below - http://buildfarm.opencsw.org/ganglia
appears to be Ganglia 3.1.3, while my packages are 3.1.7, so it seems
highly likely to be a legacy install

If you delete the old gmetad.conf, gmond.conf and httpd-ganglia.conf,
pkgrm 3.1.3 and pkgadd the new packages, it should work instantly using
multicast mode.


On 27/11/11 23:03, Daniel Pocock wrote:
> On 27/11/11 20:04, Dagobert Michelsen wrote:
>> Hi Daniel,
>>
>> Am 26.11.2011 um 04:05 schrieb Daniel Pocock:
>>> Could you please consider:
>>>
>>> a) installing gangliaagent on one or more build farm machines for
>>> monitoring?  (You need to install the gangliaweb package on an Apache
>>> machine to view the metrics, and gangliaagent on all machines to be
>>> monitored)
>>
>> Done on unstable10*
> 
> I can see it running on unstable10x, but on unstable10s,
> 
> svcs | grep gmon
> 
> shows nothing.
> 
> Also, which machine will run the web interface (CSWgangliaweb)?  I see a
> list of the PHP scripts if I navigate to
>   http://buildfarm.opencsw.org/ganglia/
> it seems that Apache is not 100% happy, it should automatically load
> index.php.  I built a fresh Solaris 10 x86 install in a VM today and
> deployed CSWgangliaweb and it worked instantly, is there possibly some
> legacy configuration in your Apache or PHP settings?
> 
> Also, the package I put out yesterday was a big broken, the ones I put
> out today are much better.
> 
>>
>>> b) installing ganglia-dev on the unstable10[sx] machines, it is needed
>>> for building other packages containing Ganglia plugins
>>
>> Done.
> 
> Great
> 
>>
>>> I am going to try and build a package that polls any kstat metric - if
>>> you look at my patch enabling PCRE in the Ganglia config, and the sample
>>> multicpu config based on PCRE, I'm sure you can imagine many ways to use
>>> it with kstat - any suggestions are welcome
>>>
>>> http://ganglia.svn.sourceforge.net/viewvc/ganglia/trunk/monitor-core/gmond/modules/conf.d/multicpu.conf?revision=2638&view=markup
>>
>> This looks very promising. Redoing the same metrics as the ones from orcallator
>> come to my mind:
>>   https://github.com/dago/orca/blob/master/data_gatherers/orcallator/orcallator.se
>>
> 
> Funny you should say that, it is exactly the same request that I got
> from the UNIX support team when I was working on Ganglia full time in
> London.
>