Re: ganglia gmond/gmetric guide?

Top Page
Attachments:
Message as email
+ (text/plain)
+ (text/plain)
Delete this message
Reply to this message
Author: der.hans
Date:  
To: Main PLUG discussion list
Subject: Re: ganglia gmond/gmetric guide?
Am 30. Mär, 2010 schwätzte Alex Dean so:

moin moin Alex,

what you wrote below is the best ganglia guide I've found. By far the best
I've found! Danke!

> On Mar 29, 2010, at 9:42 PM, der.hans wrote:
>
>> moin moin,
>>
>> I'm using ganglia for something. Someone else set it up, but I need to add
>> a bunch of metrics ( that will then be alerted via nagios ).
>>
>> Just using gmetric as is given in multiple examples isn't working.
>>
>> gmetric --name distribution_release --value $( lsb_release -d | sed -re
>> 's/Description:[[:space:]]*//' ) --type 'string'
>>
>> Running that via sudo also doesn't show up in the ganglia display.
>
> Try tracing your metric from gmond to gmetad to the web. The path it takes
> will vary depending on whether you use multicast or unicast, of course.
>
> After you run the gmetric command, wait a few seconds and run 'telnet
> localhost 8649'. gmond will give you an XML representation of all the
> metrics it knows about. You should see your new metric there.


We're on different ports, but I'm working it out.

We have different ports for different clusters/groupings. Some boxen
appear to be in multiple groups.

> On the node running gmetad, you can run 'telnet localhost 8651' to get data
> for the entire cluster. Again, your metric should be there, assuming you've
> waited long enough for gmetad to poll gmond for new data.
>
> gmetad also provides an 'interactive' port, which by default is 8652. You
> can send it a query string (which sorta looks like xpath, but isn't) and get
> the values just for a single host (or a single metric on a single host).
> This is a good path to ganglia/nagios integration.
>
> $ telnet localhost 8652
> Connected to localhost.
> Escape character is '^]'.
> /cluster_name/host_name/load_five/
>
> ... this will return just the load_five metric on host_name in the cluster
> cluster_name.
>
> If you find your metric in gmond, and in gmetad, then it sounds like you
> might have a permissions problem with the rrd files. But I seem to recall
> those kinds of problems *do* show up in /var/log/messages.


I'm certain we do have that problem.

>> Both give exit status of 0.
>>
>> It's not showing up in /var/ganglia/rrds.
>>
>> Who should own those rrds files? We have them owned by root, but gmond
>> runs as an unprivileged user?
>
> They need to be writable by the user running gmetad (nobody?), and readable
> by the Apache user.


Does ganglia automagically create the rrd files? User running gmetad needs
write perms for the rrd dir?

Thanks for the info!

ciao,

der.hans
-- 
#  http://www.LuftHans.com/Classes        http://www.TwoGeekTechs.com/
#  Director of Engineering, FonWallet Transaction Solutions, Inc.
#  "You go to Afghanistan and you swallow enough dust that you'll pass an
#  adobe brick." -- Robin Williams, 03Aug2006
---------------------------------------------------
PLUG-discuss mailing list -
To subscribe, unsubscribe, or to change your mail settings:
http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss