Graph multiple unknown named RRD values in one graph
Hi all,
I have spent the last day reading documentation and searching the "tubes" for an answer without any luck. Here is my scenario.
I have multiple hosts reporting JVM free memory numbers in the following format. Each line is a different JVM name with its current free memory number.
mn17camp.web: 846 mn18camp.web: 669 mn19camp.web: 337 mn30camp.web: 938 mn31camp.web: 750 mn32camp.web: 812 mn33camp.web: 723 mn5camp.web: 785 mn63camp.web: 179 mn65camp.web: 518 mn6camp.web: 597 mp3camp.web: 673 tbo1camp.web: 461 tbo2camp.web: 1030 tbo3camp.web: 490
I have hobbit saving them to an rrd file using the data channel in the ncv format. Is there a way to graph these without explicitly defining each jvm name? i.e. Can I iterate through whatever names are in the file? Or can I have hobbit create different rrd files based on content of the ncv file? I'm guessing it could be similar to how the disk test does it.
Thanks,
-- Michael S. Hobbs sr. systems administrator
Phone: 480.289.6832 | Fax: 480.807.0665 michael.hobbs at infusionsoft.com | www.infusionsoft.com
[cid:3301557926_12416683]
I am really rusty on dealing with NCV stuff..I know there is a way to do what you want, but not sure offhand of the proper syntax and whatnot.
As you suspect, look at the disk definition in hobbitgraph.cfg. Note that it is basically using variables for the DEF, LINE2, and GPRINT options. There is also the use of a wildcard in the rrd name definition, which comes in handy if there are multiple rrd files you are dealing with, in which case you would use: FNPATTERN (.*).web
However, from your description it sounds like you probably have a single rrd file that has multiple DS entries in it, which I have never dealt with.
Looking at the [vmstat] definition, I can see an example of pulling multiple values from a single rrd file, but you want to do it dynamically, using a variable for the DS instead of static definitions...hmm
I just had a look at some of the source code (hobbitgraph.c), and so far it looks to me like that feature is just not implemented. :(
Unless Henrik or someone can reveal the way to do it, it sounds like you are stuck with defining static DS names. Have a look at the [vmstat] definition in hobbitgraph.cfg, which should give you a good idea of how to build your own [jvm] graph definition, by using multiple DEF entries which specific the explicit DS names, example: DEF:jvm_mn17=jvm.rrd:jvm_mn17:AVERAGE DEF:jvm_mn18=jvm.rrd:jvm_mn18:AVERAGE DEF:jvm_mn19=jvm.rrd:jvm_mn19:AVERAGE
P.S. The man page for "rrdcreate" gives some great insight into the structure and use of rrd files. P.S.S. You could also consider using something external like Cacti to create and display those particular graphs.
-Charles
On Thu, Aug 14, 2008 at 11:25 AM, Michael Hobbs < michael.hobbs at infusionsoft.com> wrote:
Hi all,
I have spent the last day reading documentation and searching the "tubes" for an answer without any luck. Here is my scenario.
I have multiple hosts reporting JVM free memory numbers in the following format. Each line is a different JVM name with its current free memory number.
mn17camp.web: 846 mn18camp.web: 669 mn19camp.web: 337 mn30camp.web: 938 mn31camp.web: 750 mn32camp.web: 812 mn33camp.web: 723 mn5camp.web: 785 mn63camp.web: 179 mn65camp.web: 518 mn6camp.web: 597 mp3camp.web: 673 tbo1camp.web: 461 tbo2camp.web: 1030 tbo3camp.web: 490
I have hobbit saving them to an rrd file using the data channel in the ncv format. Is there a way to graph these without explicitly defining each jvm name? i.e. Can I iterate through whatever names are in the file? Or can I have hobbit create different rrd files based on content of the ncv file? I'm guessing it could be similar to how the disk test does it.
Thanks,
-- Michael S. Hobbs sr. systems administrator
Phone: 480.289.6832 | Fax: 480.807.0665 michael.hobbs at infusionsoft.com | www.infusionsoft.com
The issue with modeling the disk definition is that it seems to automagically create separate rrd files but reports them as a single test. It seems I would need to send a different $COLUMN for each JVM name in order to have hobbit create multiple rrd files with a single dataset in them. Is this correct? If so, then wouldn't I also need to define each $COLUMN's NCV entry in hobbitserver.cfg? It seems no matter how hobbit allows me to do it, I need to explicitly define each JVM name. Have I reached the end of hobbit's graphing abilities or am I just going about this the wrong way?
-Michael
On 8/18/08 2:13 AM, "Charles Jones" <charles.r.jones at gmail.com> wrote:
I am really rusty on dealing with NCV stuff..I know there is a way to do what you want, but not sure offhand of the proper syntax and whatnot.
As you suspect, look at the disk definition in hobbitgraph.cfg. Note that it is basically using variables for the DEF, LINE2, and GPRINT options. There is also the use of a wildcard in the rrd name definition, which comes in handy if there are multiple rrd files you are dealing with, in which case you would use: FNPATTERN (.*).web
However, from your description it sounds like you probably have a single rrd file that has multiple DS entries in it, which I have never dealt with.
Looking at the [vmstat] definition, I can see an example of pulling multiple values from a single rrd file, but you want to do it dynamically, using a variable for the DS instead of static definitions...hmm
I just had a look at some of the source code (hobbitgraph.c), and so far it looks to me like that feature is just not implemented. :(
Unless Henrik or someone can reveal the way to do it, it sounds like you are stuck with defining static DS names. Have a look at the [vmstat] definition in hobbitgraph.cfg, which should give you a good idea of how to build your own [jvm] graph definition, by using multiple DEF entries which specific the explicit DS names, example: DEF:jvm_mn17=jvm.rrd:jvm_mn17:AVERAGE DEF:jvm_mn18=jvm.rrd:jvm_mn18:AVERAGE DEF:jvm_mn19=jvm.rrd:jvm_mn19:AVERAGE
P.S. The man page for "rrdcreate" gives some great insight into the structure and use of rrd files. P.S.S. You could also consider using something external like Cacti to create and display those particular graphs.
-Charles
On Thu, Aug 14, 2008 at 11:25 AM, Michael Hobbs <michael.hobbs at infusionsoft.com> wrote: Hi all,
I have spent the last day reading documentation and searching the "tubes" for an answer without any luck. Here is my scenario.
I have multiple hosts reporting JVM free memory numbers in the following format. Each line is a different JVM name with its current free memory number.
mn17camp.web: 846 mn18camp.web: 669 mn19camp.web: 337 mn30camp.web: 938 mn31camp.web: 750 mn32camp.web: 812 mn33camp.web: 723 mn5camp.web: 785 mn63camp.web: 179 mn65camp.web: 518 mn6camp.web: 597 mp3camp.web: 673 tbo1camp.web: 461 tbo2camp.web: 1030 tbo3camp.web: 490
I have hobbit saving them to an rrd file using the data channel in the ncv format. Is there a way to graph these without explicitly defining each jvm name? i.e. Can I iterate through whatever names are in the file? Or can I have hobbit create different rrd files based on content of the ncv file? I'm guessing it could be similar to how the disk test does it.
Thanks,
-- Michael S. Hobbs sr. systems administrator
Phone: 480.289.6832 | Fax: 480.807.0665 michael.hobbs at infusionsoft.com | www.infusionsoft.com
[cid:3301895082_32644645]
I think you need the separate RRD file for each element you want to graph, with a naming convention. Then the hobbitgraph definition can "wildcard" the file names to put the graphs together. This is what is done for disk graphs, and network interface graphs, etc.
I know nothing about NCV, but it is pretty easy to do this if you write your own script for processing custom tests. Then you can control how each RRD is created, and enforce your own naming convention. The trick is to thoroughly understand how the current "multi-graph" tests work, and then try to duplicate them with different data.
GLH
From: Michael Hobbs [mailto:michael.hobbs at infusionsoft.com]
Sent: Monday, August 18, 2008 11:05 AM
To: hobbit at hswn.dk
Subject: Re: [hobbit] Graph multiple unknown named RRD values in
one graph
The issue with modeling the disk definition is that it seems to
automagically create separate rrd files but reports them as a single test. It seems I would need to send a different $COLUMN for each JVM name in order to have hobbit create multiple rrd files with a single dataset in them. Is this correct? If so, then wouldn't I also need to define each $COLUMN's NCV entry in hobbitserver.cfg? It seems no matter how hobbit allows me to do it, I need to explicitly define each JVM name. Have I reached the end of hobbit's graphing abilities or am I just going about this the wrong way?
-Michael
On 8/18/08 2:13 AM, "Charles Jones" <charles.r.jones at gmail.com>
wrote:
I am really rusty on dealing with NCV stuff..I know
there is a way to do what you want, but not sure offhand of the proper syntax and whatnot. As you suspect, look at the disk definition in hobbitgraph.cfg. Note that it is basically using variables for the DEF, LINE2, and GPRINT options. There is also the use of a wildcard in the rrd name definition, which comes in handy if there are multiple rrd files you are dealing with, in which case you would use: FNPATTERN (.*).web However, from your description it sounds like you probably have a single rrd file that has multiple DS entries in it, which I have never dealt with. Looking at the [vmstat] definition, I can see an example of pulling multiple values from a single rrd file, but you want to do it dynamically, using a variable for the DS instead of static definitions...hmm I just had a look at some of the source code (hobbitgraph.c), and so far it looks to me like that feature is just not implemented. :( Unless Henrik or someone can reveal the way to do it, it sounds like you are stuck with defining static DS names. Have a look at the [vmstat] definition in hobbitgraph.cfg, which should give you a good idea of how to build your own [jvm] graph definition, by using multiple DEF entries which specific the explicit DS names, example: DEF:jvm_mn17=jvm.rrd:jvm_mn17:AVERAGE DEF:jvm_mn18=jvm.rrd:jvm_mn18:AVERAGE DEF:jvm_mn19=jvm.rrd:jvm_mn19:AVERAGE P.S. The man page for "rrdcreate" gives some great insight into the structure and use of rrd files. P.S.S. You could also consider using something external like Cacti to create and display those particular graphs. -Charles On Thu, Aug 14, 2008 at 11:25 AM, Michael Hobbs < michael.hobbs at infusionsoft.com> wrote:
Hi all,
I have spent the last day reading documentation
and searching the "tubes" for an answer without any luck. Here is my scenario. I have multiple hosts reporting JVM free memory numbers in the following format. Each line is a different JVM name with its current free memory number. mn17camp.web: 846 mn18camp.web: 669 mn19camp.web: 337 mn30camp.web: 938 mn31camp.web: 750 mn32camp.web: 812 mn33camp.web: 723 mn5camp.web: 785 mn63camp.web: 179 mn65camp.web: 518 mn6camp.web: 597 mp3camp.web: 673 tbo1camp.web: 461 tbo2camp.web: 1030 tbo3camp.web: 490 I have hobbit saving them to an rrd file using the data channel in the ncv format. Is there a way to graph these without explicitly defining each jvm name? i.e. Can I iterate through whatever names are in the file? Or can I have hobbit create different rrd files based on content of the ncv file? I'm guessing it could be similar to how the disk test does it. Thanks,
--
Michael S. Hobbs
sr. systems administrator
Phone: 480.289.6832 | Fax: 480.807.0665
michael.hobbs at infusionsoft.com | www.infusionsoft.com
participants (3)
-
charles.r.jones@gmail.com
-
greg.hubbard@eds.com
-
michael.hobbs@infusionsoft.com