Archive for the 'Development' Category

Passing key value pairs through kernel messages with libnagios

If you caught my last post, you learned about compiling libnagios and using the Nagios library in your own code. This time we’re going to use that same Nagios library along with some borrowed NDOUtils code and build something super neat: a message passing system utilizing key-value vectors and buffers attached to the kernel messaging (IPC) utilities from NDO.

Just like last time, I’m going to assume you have a sane build environment set up (where tools like make and ./configure are working) before we go any further. If you are following along, now would be the time to get these in order.

Before we get started writing our own code, we need to borrow some files from NDOUtils. We only need two files: src/queue.c and include/queue.h. You can find the source for each at and, respectively. Once you have them downloaded, we’ll be ready to move on.

First, open up queue.c in your favorite editor and delete the line that includes “../include/config.h” (we’ll manually include the headers we need in a bit). You’ll want to change the path to your queue.h file to reflect wherever you downloaded it. For the purposes of this post and the examples that follow, all of my source and header files will be in the root directory of the Nagios Core source code that we downloaded while following along with the last post.

So, the top of your queue.c should have looked like this when you downloaded it:

After you’ve made the changes I mentioned, it should look like this:

Save that file and open up queue.h. We’re going to add all of the necessary header files that would have been included with config.h in here. Find the line that defines NDOUTILS_INCLUDE_QUEUE_H, and directly after it add the following include directives:

By now, the top of your queue.h file should look like this:

Perfect! Now we’re going to set up our main application. Last time I walked through the file as we were creating it, but this time it’s a tad more advanced so we’re going to look at it in its entirety and then step through the explanation. Create a file named test2.c, and fill it with the following content:

Here we’re just including our necessary header files:

This is a function that we use to “walk” through the key-value vector and perform some meaningful operation against each pair we find. *kv will be the key-value pair, and *arg is an additional argument you can pass to the kvvec_foreach function (

Now we’re going to start our application and define a few parameters to be used later. KEYVAL_SEP will be the separating character used to determine key from values. PAIR_SEP will be the separating character to keep key-value pairs separate from each other. OVERALLOC will be used to determine how much extra space is allocated when we create our vector. Then we go ahead and set up a basic char * buffer with our key-value data and set keyval_len equal to how many [key-values] we have there.

Now we’re getting to the nitty gritty. It’s time to declare some vectors and a buffer. We’ll initialize one for use immediately, but save the others for later. We’re using kvvec and kvvec_buf structures to store vectors and buffers ( and initializing an empty kvvec with kvvec_create (

Let’s cycle through our *keyval array and inject our pairs into our freshly initialized kvvec using kvvec_addkv (

Remember the function we defined that was going to “walk” through the key-value vector and perform a meaningful operation against each key-value pair? We’re going to call it now, and as it walks through each key-value, it will print both elements.

Now we’re going to sort the vector using kvvec_sort (, which will sort the pairs alphabetically by key. Once we’ve done that, we’re going to call the walker function again which will print the key-value pairs in their new order.

The magic and beauty of libnagios’s kvvec function really shines through with the buffer/vector conversion functions. We’re going to use kvvec2buf ( to convert our vector into a single buffer using the parameters we set up when we began our application.

Now, we finally get to the borrowed NDOUtils code. This next block is responsible for initializing a message queue utilizing ndo2db_queue_init ( It takes only one argument: an integer ID to be used as identification for the message queue. We’re passing our process id.

Next we’re creating a struct ndo2db_queue_msg ( variable and copying our key-value buffer.

Finally, we’ll send our message to the queue using ndo2db_queue_send ( This function takes two arguments: an ndo2db_queue_msg and a size_t the size of the message to send.

Our example here is very basic. Ideally, it is at this point in the application that something would happen to this buffer. It would be picked up by some child process or thread or a different process entirely and processed there. In order to maintain some brevity, we’ll simply be pulling it right back out of the queue! In order to do that, we’re using pop_from_queue ( This simply returns a char pointer containing your message text (which happens to be the buffer text). Once we’ve done that, we’re going to use buf2kvvec ( to convert our message passed buffer back into a key-value vector.

After that conversion takes place, we’ll issue a call to the walker function yet again just so we can make sure that our keys and values are right where we expect them to be.

Whew. How exciting! All that’s left now is to clean up the mess we made. In order to do that, we need to call ndo2db_queue_free ( so that we release the queue’s system resources. After that, we’re freeing some buffers and then destroying the key-value vectors using kvvec_destroy ( We’re passing the KVVEC_FREE_ALL flag here, but you could decide to only KVVEC_FREE_KEYS or KVVEC_FREE_VALUES.

Just in case you missed it last time, we’ll need to have libnagios compiled and installed if we want any of this to work. If you haven’t already, you can do this with the following commands:

Now that we have libnagios all set up and usable, we can compile our test2 application. You can do that with the following command:

Now lets run our application and see the output!

Neat! Just as a side note, when you’ve pushed your messages onto the kernel message queue, you can actually watch the queue with the command ipcs -q.

I hope you’ve learned a little more about using the Nagios library to extend your own code.

Extending Nagios functionality with libnagios

Have you ever attempted to write a function in C to execute a command and parse the output? I think I’d rather just let the Nagios library do the heavy lifting for me.

This blog post is going to cover the basics of compiling libnagios, and linking the Nagios library to your application. I’ll be focusing on using some of the built-in Nagios functionality, specifically the runcmd_open() function.

I’m going to assume you have a sane build environment set up (where tools like make and ./configure are working) before we go any further. If you are following along, now would be the time to get these in order.

First, download the source code and extract it. You can get a copy of the Nagios Core source at Once you’ve downloaded it and extracted the files, open up nagioscore-master/lib/runcmd.h ( Search for “extern int runcmd_open”, as of the time of this writing, that should bring you to line 77, where our function is declared:


So what does all that mean? It means we need a command to execute, a file descriptor for stdout, another filedescriptor for stderr. Our application doesn’t need a callback function to register iobrokers or a value to pass. But, since these are declared non null, we’ll have to get creative.

Let’s create a file, named test.c in the root of the nagioscore-master directory. First, we need to include our libnagios header.


Then we define our fake iobroker_register function. This is essentially just a placeholder, as we aren’t (yet) particularly interested in assigning a function to execute when our stdout/stderr stops reading.


Next, we set up our variables that we’ll be using to pass to the runcmd_open() function. We don’t need an env variable, since that argument can accept a NULL value, we’re just going to pass that in (especially since it is unused anyway).


Now we execute runcmd_open(), and let the Nagios library do its magic! This will put stdout in pfd[0] and stderr in pfderr[0].


Let’s copy the stdout to our out var and print some information relating to the command we executed and that command’s output.


Finally, we clean up our memory and exit the program.


Here’s the file in its entirety:



Let’s see it in action! First we’re going to compile our Nagios library! Open up your terminal and let’s get to library compilin’:


Those commands should have compiled your Nagios library and then placed it in /usr/local/nagios/lib. Now, we’re finally ready to compile our program:


Now, if everything went well up this point, you should be able to execute our basic program with the following command:

Your output should be similar to the following:

I hope that you’ve learned a few things about using the Nagios library in your own code. Questions, comments, and suggestions for future posts are all welcome below in the comments section.


– Bryan Heden

Exploring the New JSON CGIs in Nagios Core 4.0.7! (Part 1)

The JSON CGIs, from the JSON branch of core, have been officially released with Nagios Core 4.0.7!

The original design goals were:

  1. To provide all information available in current CGIs in JSON format.
  2. Place the presentation responsibility on the client to minimize network traffic and server load.
  3. Perform operations on the server side that are significantly more easily done there.
  4. Spark community developers to create new Nagios Core UI’s from the easy to work with JSON from the CGIs.

The CGIs provide an API to query object, status, and historical information through GET requests.  They use the same authentication as other CGIs.  Once queried, they return valid JSON that can be parsed into JavaScript objects for client side models and processing.  The API is very robust, providing multiple ways to limit queries – name/descriptions, host/service groups,  update/changes times, among many others.

The three new CGIs are:

  1. objectjson.cgi  (object configuration)
  2. statusjson.cgi  (status information)
  3. archivejson.cgi  (historical logs)

Additionally, a new web app is included – jsonquery.html & jsonquery.js.  This is a small UI for crafting GET requests, it can be used to trial specific parameters for GET requests, or to just explore the api.  It is also the easiest way to get acquainted with the new CGIs.

Continue reading ‘Exploring the New JSON CGIs in Nagios Core 4.0.7! (Part 1)’

Major Improvements to Agent-Based Monitoring with the Release of Nagios Cross-Platform Agent — NCPA

Major improvements to agent-based monitoring have been taking place at Nagios Enterprises. NCPA, the Nagios Cross-Platform Agent, is a project that has the potential to revolutionize agent-based monitoring and increase the efficiency of IT support teams world-wide.

As many Nagios users know, monitoring with agents means juggling the installation of many different types of plugins to try and match devices, operating systems, and the basic functions of each agent. For example, in a simple agent-based Linux and Windows server environment you have to install 2 agents, know the 2 user manuals, there are 2 times the troubleshooting hours required, 2 times the commands on remote systems, 2 change logs to sift through for potential update breaks…the list goes on. It can be very difficult to keep organized and take a lot of time to implement and update your configuration, especially when your monitoring environment becomes larger and more complex.

Whether your environment is large or small, there are usually a myriad of devices that need to be monitored and more often than not, some sort of agent needs to be installed on these devices.

Wouldn’t it be simple if you only had to install one agent regardless of operating system or device?

We have been working on a project that aims to do this. Nagios Cross-Platform Agent (NCPA) is a fully contained agent that runs on Mac OS X, Windows, and Linux and seeks to solve all of the previously mentioned pitfalls of agent based monitoring with Nagios. The main goal of NCPA was to monitor the core metrics of a server and other devices without the added hassle of plugins and dependencies. Metrics such as CPU Usage, Disk Usage, Memory Usage, Interface Usage, Swap Usage, User Count, etc. are preloaded in NCPA so that all you have to do is install the agent. It has since broadened in scope to be a general purpose agent that is very good at doing the aforementioned job. Just install the NCPA agent on your system, and away you go.

Features & Benefits of NCPA:

-Installs on multiple platforms : Windows, Linux, Mac OS X and FreeBSD (untested on AIX, HPUX and Solaris)
-Real-time performance graphs and GUI configuration
-Fully contained agent, including dependencies
-Identical cross-platform configuration editing for both active and passive agents
-Minimizes knowledge needed to know down to one
-Advanced visual data representation

Direct links to the NCPA .exe and .rpm files can be found in the installation instructions which can be downloaded at the link below: Installing NCPA.pdf

We are very excited about this new agent and are currently looking for real world testers to try it out. To test NCPA please contact Thanks!