NCV variable for custom RRD's
list Ivan Verbeeck
Hello all, I recently added a custom test to our hobbit setup that will use the NCV format for the rrd graph. The only problem that i noticed is the option to use wildcards in the NCV_test variable to define the "default" dataset type. When i specify all the datasets in the environment variable and set them to GAUGE (wich i need for my test), it all works fine, but if i use the "*:GAUGE" in this variable the rrd is generated with datasets of the "DERIVE" type. Anyone a idea what can be wrong? Kind regards, Ivan
list Henrik Størner
▸
On Mon, May 15, 2006 at 12:01:43PM +0200, Ivan Verbeeck wrote:
I recently added a custom test to our hobbit setup that will use the NCV format for the rrd graph. The only problem that i noticed is the option to use wildcards in the NCV_test variable to define the "default" dataset type. When i specify all the datasets in the environment variable and set them to GAUGE (wich i need for my test), it all works fine, but if i use the "*:GAUGE" in this variable the rrd is generated with datasets of the "DERIVE" type. Anyone a idea what can be wrong?
The wildcard thing is only available in the 4.2 alfas or the latest snapshot. If you're using 4.1.x then it is not implemented there. Regards, Henrik
list Ivan Verbeeck
Henrik, Thanks for this info, i'm still using 4.1.2, so i know how to fix it. I still have two questions about this NCV system : - Is it possible to define a graph in hobbitgraph.cfg for rrd files with "variable" datasets. By example when i send status messages for different temperature sensors of a network switch (chassis, power supply, ...) the data in the (single) rrd looks good whith the different ds names. So if it is possible to just graph all the existing datasets in the rrd, i can use a single rrd-config for the temperature of different switchen, servers, ..., otherwise i need a different "temp"-testname for all the different devices. - How can i send graph data to hobbit for a "cpu" test for switches. Can this send by using a status message, or do i need to send a data message (like hobbitclient)? If it must be with a data message, i found in de man pages "data HOSTNAME.DATANAME<newline><additional text>", what is the DATANAME for the cpu data, and is there a minimum data that must be send ([date],[uname],[uptime],..). I know i can just rename the "cpu" test to something else, but "cpu" says it all :-) Kind regards, Ivan
▸
On Mon, May 15, 2006 at 12:01:43PM +0200, Ivan Verbeeck wrote:I recently added a custom test to our hobbit setup that will use the
NCV
format for the rrd graph. The only problem that i noticed is the option to use wildcards in the NCV_test variable to define the "default" dataset type. When i specify all the datasets in the environment variable and set them to GAUGE (wich i need for my test), it all works fine, but if i use the "*:GAUGE" in this variable the rrd is generated with datasets of the "DERIVE" type. Anyone a idea what can be wrong?
The wildcard thing is only available in the 4.2 alfas or the latest snapshot. If you're using 4.1.x then it is not implemented there. Regards, Henrik
list Henrik Størner
▸
On Fri, May 26, 2006 at 11:11:30PM +0200, Ivan Verbeeck wrote:
I still have two questions about this NCV system : - Is it possible to define a graph in hobbitgraph.cfg for rrd files with "variable" datasets. By example when i send status messages for different temperature sensors of a network switch (chassis, power supply, ...) the data in the (single) rrd looks good whith the different ds names. So if it is possible to just graph all the existing datasets in the rrd, i can use a single rrd-config for the temperature of different switchen, servers, ..., otherwise i need a different "temp"-testname for all the different devices.
Currently the NCV system bundles all datasets into one RRD file, and therefore you need one graph definition for each of the dataset combinations you will get. That is obviously not very smart for the kind of data you mention. It would be nice to have some option to split it into one dataset per RRD file - like what Hobbit does for disk graphs - but that is not possible right now.
▸
- How can i send graph data to hobbit for a "cpu" test for switches. Can this send by using a status message, or do i need to send a data message (like hobbitclient)? If it must be with a data message, i found in de man pages "data HOSTNAME.DATANAME<newline><additional text>", what is the DATANAME for the cpu data, and is there a minimum data that must be send ([date],[uname],[uptime],..). I know i can just rename the "cpu" test to something else, but "cpu" says it all :-)
You can send it as "status" or "data" - but if you want it to show up on the webpage as a status-column, you must use "status". If you want to track the cpu utilisation in a graph, the simplest way is if you can format the status message so it looks like a status message from either a Unix- or a Windows-client; then the graph data will automatically be picked up by Hobbit. If that is not possible, then things begin to get difficult. You would have to write your own Hobbit server module to pick up the status messages, and have it parse the "cpu" reports you send into it like the "hobbitd_rrd" task currently does. Regards, Henrik