Skip to main content

 In the observability era, many metrics are collected on various devices. Sometimes, metrics gathered from a Centreon Host are even more interesting when compared to or associated with others.

This tutorial will guide you through creating a single check aggregating metrics from multiple Hosts and Services. 

The Centreon SQL Metrics Plugin Pack offers a handy service template: App-Monitoring-Centreon-SQL-Virtual-Curves-custom. This template will help you configure such a check. 

 

The idea behind it is to offer an alternative to the good old meta services which have several limitations, as an example they: 

  • cannot make the output human-readable
  • cannot do advanced calculations on multiple metrics
  • cannot be used efficiently in Business edition modules.

 

Concepts

 

This screenshot of a detailed Service page highlights the main monitoring concepts you must understand to avoid getting lost during this tutorial. Pretty simple right?
 

 

Use cases

 

Here are some examples of what we've done so far with our community users and customers:

  • Sum different network devices bandwidth utilization (example: incoming or outgoing traffic of a specific network zone)
  • Sum storage utilization (Space used or provisioned across all your VMWare datastores) 
  • Get average, min, max of individual Services' metrics (example: connected users or license usage)
  • Perform specific mathematical operations on metrics to create new ones (example: Convert power usage into CO2 emanation)
  • Display within the same service or graphs curves generated by several others

In this tutorial, we will see how we can sum the total bandwidth going through the data interface of both nodes of my proxy cluster users configure to access internet:

 

How it works?

 

Before going deeper in the use case, you probably wonder how it works behind the scene.

The plugin connects to the Centreon database and reads a configuration file to perform advanced computing on metrics. 

It uses the database to fetch metrics' current values during its execution.

The configuration file defines:

  • Which Hosts, Services, Metrics the plugin should consider?
  • How to mould raw metrics?
  • How to output the result?

Configuration file 

 

Here is a complete configuration file to get a global idea of the layout. We will then look closer at each part and how directives enable some plugin features.

 

{
"selection": {
"traffic_in_node_1": {
"host_name": "HQ-PXY-Inet-1",
"service_name": "Traffic-data",
"metric_name": "traffic_in",
"display": false
},
"traffic_in_node_2": {
"host_name": "HQ-PXY-Inet-2",
"service_name": "Traffic-data",
"metric_name": "traffic_in",
"display": false
},
"traffic_out_node_1": {
"host_name": "HQ-PXY-Inet-1",
"service_name": "Traffic-data",
"metric_name": "traffic_out",
"display": false
},
"traffic_out_node_2": {
"host_name": "HQ-PXY-Inet-2",
"service_name": "Traffic-data",
"metric_name": "traffic_out",
"display": false
}
},
"virtualcurve": {
"sum_traffic_in": {
"pattern": "^traffic_in$",
"aggregation":"sum"
},
"sum_traffic_out": {
"pattern": "^traffic_out$",
"aggregation":"sum"
}
},
"formatting": {
"change_bytes_network": 1,
"custom_message_global": "Traffic on data interfaces is OK",
"printf_msg": "Incoming traffic SUM is: %s %s/s",
"printf_var": "$self->{result_values}->{value}, $self->{result_values}->{unit}"
}
}

 

Basic use

 

The "selection" JSON node is meant to describe metrics to be selected: 

  •  "traffic_*node_1" & "traffic_*node_2" are selection identifiers. Both have their keys:
    • "host_name": The exact hostname in Centreon
    • "service_name": The exact service name in Centreon
    • "metric_name": The metric name (you can use a POSIX regex filter)
    • "display": Define if this metric should create a dedicated curve.

The "virtualcurves" JSON node is meant to describe the new curves to add  created:

  • "sum_traffic_in" & "sum_traffic_out" are virtual curves identifiers. Both have their keys:
    • "pattern": Pattern to process a part of previously selected metrics inside the virtual curve. Here we use a regex to match exactly a metric name. 
    • "aggregation": Aggregation you want to apply. Possible values are: 'min', 'max,' 'sum', 'avg' or 'none'. Here we will sum value from the two metrics matching the pattern. 
    •  

The “formatting” sub-section defines the plugin output

  • "change_bytes_network": Use base 1000 to convert metric result to a human-readable format. You can use change_bytes instead if you deal with storage volumes. 
  • "printf_msg": Printf expression, if you want to handle advanced data type you may want to read this but it will work with ‘%s’ to substitue value as a string
  • "printf_var": Scalar to use within Printf substitution. Stick to the example as they are the only ones available at the moment. If you don’t need to set change_bytes to true, simply remove “, $self->{result_values}->{unit}”.

Advanced use

 

The "selection" JSON node may be substituted by a "filters" one. It allows SQL '%' wildcard utilization and makes the configuration much simpler. 
 

        "filters": {
                "host": "HQ-PXY-Inet%",
                "service": "Traffic-data",
                "metric": "traffic_%",
                "display": false
        },

Please note the differences in the name of the keys:
* "host": Hostname filter
* "service": Service filter
* "metric": Metric filter

 

Setup and practice 

 

Prerequisites

 

The configuration file must be readable by the centreon-engine user. 

The Poller executing the check must be able to connect to the centreon_storage database over 3306/TCP port with values supplied through --username and --password options.

The SQL user must hold required privileges to "SELECT" data within index_data and metrics tables part of the centreon_storage database.

 

Installation 

Create a directory to store your config files:

mkdir /etc/centreon-engine/virtual-curve-cfg/

Set the centreon-engine user as owner of this directory 

chown centreon-engine. /etc/centreon-engine/virtual-curve-cfg/

Install the Plugin Pack: 

  • Install the Plugin code on the Central or/and a Poller
yum install centreon-plugin-Applications-Monitoring-Centreon-SQL-Metrics

2. Install the 'Centreon SQL Metrics" Plugin Pack through "Configuration > Plugin packs > Manager" page. Note that if you use an offline licence you will need to issue this command on your central server to see it listed: 

yum install centreon-pack-applications-monitoring-centreon-sql-metrics

 

Plugin execution 

 

You can execute the plugin like this:

/usr/lib/centreon/plugins/centreon_centreon_sql_metrics.pl \ 
--plugin=database::mysql::plugin \
--dyn-mode=apps::centreon::sql::mode::virtualservice \
--host='127.0.0.1' \
--username='<your-db-user>' \
--password='<your-db-passwd>' \
--config-file=/etc/centreon-engine/virtual-curve-cfg/proxy-sum-data.json' \
--verbose

Here is the result of the previous execution: 

OK: Traffic on data interfaces is OK | 'sum_traffic_in'=102394;;;; 'sum_traffic_out'=66941;;;;
Incoming traffic SUM is: 102.39 Kb/s
Incoming traffic SUM is: 66.94 Kb/s

As every other plugin, you can add the ```--help``` flag to see all available options. 

It's possible to set WARNING and CRITICAL thresholds with these options:

--warning-metric and --critical-metric

 

--warning-global and --critical-global

 

Going further and how modifications on the config reflect on the plugin behavior?

 

Let's modify the configuration file and see what it changes.

  • Set display as true in your filters section. Note that all selected metrics are now visible in the verbose output:  

Your "filters" section then looks like this: 

        "filters": {
                "host": "HQ-PXY-Inet%",
                "service": "Traffic-data",
                "metric": "traffic_%",
                "display": true
        },

Plugin execution now produces the output shown below and display details for each metrics: 

OK: Traffic on data interfaces is OK - All metrics are OK | 'sum_traffic_in'=113765;;;; 'sum_traffic_out'=63588;;;; 'HQ-PXY-Inet-1#Traffic-data#traffic_in'=105291b/s;;;0;1000000000 'HQ-PXY-Inet-1#Traffic-data#traffic_out'=53946.2b/s;;;0;1000000000 'HQ-PXY-Inet-2#Traffic-data#traffic_in'=8474.08b/s;;;0;1000000000 'HQ-PXY-Inet-2#Traffic-data#traffic_out'=9642b/s;;;0;1000000000
Incoming traffic SUM is: 113.77 Kb/s
Incoming traffic SUM is: 63.59 Kb/s
Metric 'HQ-PXY-Inet-1#Traffic-data#traffic_in' value is '105.29'
Metric 'HQ-PXY-Inet-1#Traffic-data#traffic_out' value is '53.95'
Metric 'HQ-PXY-Inet-2#Traffic-data#traffic_in' value is '8.47'
Metric 'HQ-PXY-Inet-2#Traffic-data#traffic_out' value is '9.64'

Adding a “formatting” subsection in the “filters” node will allow to display the traffic of each interface in a readable format:

OK: Traffic on data interfaces is OK - All metrics are OK | 'sum_traffic_in'=105095;;;; 'sum_traffic_out'=52581;;;; 'HQ-PXY-Inet-1#Traffic-data#traffic_in'=104250b/s;;;0;1000000000 'HQ-PXY-Inet-1#Traffic-data#traffic_out'=44157b/s;;;0;1000000000 'HQ-PXY-Inet-2#Traffic-data#traffic_in'=845.74b/s;;;0;1000000000 'HQ-PXY-Inet-2#Traffic-data#traffic_out'=8424.66b/s;;;0;1000000000
Incoming traffic SUM is: 105.09 Kb/s
Incoming traffic SUM is: 52.58 Kb/s
'HQ-PXY-Inet-1#Traffic-data#traffic_in' bandwidth is: 104.25 Kb/s
'HQ-PXY-Inet-1#Traffic-data#traffic_out' bandwidth is: 44.16 Kb/s
'HQ-PXY-Inet-2#Traffic-data#traffic_in' bandwidth is: 845.74 b/s
'HQ-PXY-Inet-2#Traffic-data#traffic_out' bandwidth is: 8.42 Kb/s

Centreon configuration 

 

Now you can configure all of this in Centreon. Add a Host and apply the relevant template. Please fill the IP with the address of Centreon database server and set a username and a password with required privileges: 

 

Then, create a service using the Virtual Curves template and fill the CONFIGJSON macro with the path to your config file: 

 

Save and export your configuration. 

 

Enjoy your virtual service! 

 

 

Hi,

It works fine, thanks 👍 !


Hello @qgarnier,

Thanks!

We are testing with this option and we will get back to you.


Ok got it. Add following option to have 0 value:

--check-metrics=''

 


Hi,

To be more explicit : 2 internet accesses in active/passive mode. The inactive access does not return an error. Its status and admin are both down, and the followig filter is embeded in the check command

--critical-status='%{admstatus} eq "up" and %{opstatus} ne "up"'

So the status of the port id down, but the check is OK.

And here is an example of what happens with the virtual curve (the last one) when access is switch :
 

Is there a way to set to 0 the value of the check which does’nt return a value anymore ?

Regards


Hello @qgarnier,

Thank you for the return.

We actually use centreon-plugins with interfaces mode.

If we understand correctly, there is no possibility to ignore a deprecated value? If one of the input values is in error, the virtual-service will therefore necessarily be false.

Regards

 


@Nico Which plugin do you use to have the traffic ? Because with centreon-plugins and mode interfaces you should have a 0 traffic value if you have an interface status error.

 

Nevertheless,  virtual-service cannot ignore an old value (because it doesn’t know that the metric value is outdated).


Allo?


Hi,
We've a tricky case to submit, here's a simplified version.
For example we've got 2 internet access, only one active at a time : one port is up with traffic while the other one is down.
We created a virtual service to display traffic curve whatever the active access is, the vr is a sum of the 2 traffic metric.
As a result, information display through this curve is always false because it sums the last value collected for both the active and inactive access.
So one can see artificial lower value in the curves :

Do you have any work around to suggest ?

Regards

 

Hi,

Have you seen any news about this particular case?

Best.


hi,

hiding database macros as password causes command to fail

/usr/lib/centreon/plugins/centreon_centreon_sql_metrics.pl --plugin=database::mysql::plugin --dyn-mode=apps::centreon::sql::mode::virtualservice --host='10.0.11.18' --username='***' --password='***'  --config-file='/etc/centreon-engine/virtual-curve-cfg/test.json' --warning-global='' --critical-global='' --warning-metric='' --critical-metric=''

UNKNOWN: Cannot connect: Access denied for user '***'@'10.0.8.115' (using password: YES)

 


Because if I try with this config :

    "filters": {
"host": "NEXUS_DATACENTER_EPERNON_PRA_01",
"service": "Traffic-data",
"metric": "traffic_%",
"display": true,
"formatting": {
"change_bytes_network": 1,
"custom_message_global": "Traffic on data interfaces is OK",
"printf_msg": "Incoming traffic SUM is: %s %s/s",
"printf_var": "$self->{result_values}->{value}, $self->{result_values}->{unit}"
}
}

At the end, I have this return :

OK: Global metrics are OK | 'sum_traffic_in'=8016;;;; 'sum_traffic_out'=5884;;;;
Metric 'sum_traffic_in' value is '8016'
Metric 'sum_traffic_out' value is '5884'

It’s better but I wanna have the readable format. How can I do ?

 

Thanks

Regards


e


Hi,

It’s OK for me but why two incoming traffic ? There is incoming and outcoming ? So how make change in script to have IN and OUT ?

OK: Traffic on data interfaces is OK | 'sum_traffic_in'=9871;;;; 'sum_traffic_out'=6384;;;;
Incoming traffic SUM is: 9.87 Kb/s
Incoming traffic SUM is: 6.38 Kb/s

 

"formatting": {
        "change_bytes_network": 1,
        "custom_message_global": "Traffic on data interfaces is OK",
        "printf_msg": "Incoming traffic SUM is: %s %s/s",
        "printf_var": "$self->{result_values}->{value}, $self->{result_values}->{unit}"
    }

 

 

Have you any idea ? Use filters with formating in ?

 

Thanks

Regards

 


By default, we use %d to display the metric value to avoid unecessary ‘.00’ at the and stick to a simple  decimal digit.

 

If you need to be more precise, you have to specify the sprintf formatting you want (https://perldoc.perl.org/functions/sprintf). 

 

%f => float

%.2f => round to two digit after the decimal point 

%.3f => round to three digit after the decimal point 

 

You’re welcome! 

 


ohhh yeah ! 😁

seems to be ok now

could you explain me ? 

Thanks a lot!  


Can you try to add: 

 

“printf_metric_value”: “%.2f” 

 

in the formatting section please?


Hi @sims24,

thanks for quick response, 

Same behavior with %.2f

 


Hi, 

Aggregate function seems to work well with integers but not with float metrics.
I’m trying to create aggregated metric for response time ex : 0.1112 0.223 etc… but the virtual curve always send me 0. 

If i use another metric without float it is ok. 

 

 

Is anyone can help me about this ? 

Thanks, 

 

Try using %.2f instead of %s in your printf message please. 


Hi, 

Aggregate function seems to work well with integers but not with float metrics.
I’m trying to create aggregated metric for response time ex : 0.1112 0.223 etc… but the virtual curve always send me 0. 

If i use another metric without float it is ok. 

 

 

Is anyone can help me about this ? 

Thanks, 

 


Hi,
We've a tricky case to submit, here's a simplified version.
For example we've got 2 internet access, only one active at a time : one port is up with traffic while the other one is down.
We created a virtual service to display traffic curve whatever the active access is, the vr is a sum of the 2 traffic metric.
As a result, information display through this curve is always false because it sums the last value collected for both the active and inactive access.
So one can see artificial lower value in the curves :

Do you have any work around to suggest ?

Regards


👍


Thank you @FredericGerard I’ll track this for the next release. 

 

Simon


Hi,

We have just installed the plugin and plugin pack. Macros CENTREONDATABASEUSER and CENTREONDATABASE are missing int the definition of the template. We’ve had to add them to make the test work properly.

Regards.


Reply