Skip to main content

Collect OpenMetrics

There is an effort to promote Prometheus exposition format into a standard known as OpenMetrics. Some products adopted the format: InfluxData's suite, InfluxDB and Google Cloud Platform.

Installation​

  1. Install the plugin on every poller expected to monitor openmetrics:

    yum install centreon-plugin-Applications-Monitoring-Openmetrics

  2. Install the "OpenMetrics" Centreon Monitoring Connector from the Configuration > Monitoring Connector Manager page.

Configuration​

The Monitoring Connector "OpenMetrics" provide the collect of openmetrics from file on the collector (the template will be App-Monitoring-Openmetrics-File-custom) or from a API providing openmetrics (the template will be App-Monitoring-Openmetrics-Web-custom). Following the template, several macros must be configured :

App-Monitoring-Openmetrics-File-custom​

MandatoryNameDescription
XOPENMETRICSFILEPATHPath of the file that containts openmetrics (eg. /var/data/metrics)
OPENMETRICSEXTRAOPTIONSAny extraoptions you may want to add to every command line (eg. a --verbose flag)

App-Monitoring-Openmetrics-Web-custom​

MandatoryNameDescription
XOPENMETRICSPORTPort use by the endpoint (eg '80', '443' or '9000')
XOPENMETRICSPROTOProtocol use by the endpoint ('http' or 'https')
XOPENMETRICSURLPATHPath of the url that expose openmetrics (eg. /metrics)
OPENMETRICSUSERNAMEUsername in case of authentication required
OPENMETRICSPASSWORDPassword in case of authentication required
OPENMETRICSEXTRAOPTIONSAny extraoptions you may want to add to every command line (eg. a --verbose flag)

Thoses 2 host templates will deploy automatically a service "Scrape-Metrics" bind to your host. This service can be duplicated as often as necessary according to the number of metrics to collect.

How To​

Collect metrics using the command line​

We have a server/container Prometheus which expose metrics ? Or a server/container Cadvisor which also expose metrics ? We are going to see some examples to collect thoses metrics from thoses endpoints.

Prometheus​

Our server/container Prometheus expose metrics from this url: <http://prometheus.int.centreon.com:9090/metrics>

# HELP prometheus_http_requests_total Counter of HTTP requests.
# TYPE prometheus_http_requests_total counter
prometheus_http_requests_total{code="200",handler="/alerts"} 1
prometheus_http_requests_total{code="200",handler="/api/v1/label/:name/values"} 24
prometheus_http_requests_total{code="200",handler="/api/v1/labels"} 3
prometheus_http_requests_total{code="200",handler="/api/v1/query"} 5454
prometheus_http_requests_total{code="200",handler="/api/v1/query_range"} 3390
prometheus_http_requests_total{code="200",handler="/api/v1/targets"} 3
prometheus_http_requests_total{code="200",handler="/config"} 2
prometheus_http_requests_total{code="200",handler="/flags"} 2
prometheus_http_requests_total{code="200",handler="/graph"} 17
prometheus_http_requests_total{code="200",handler="/metrics"} 155749
prometheus_http_requests_total{code="200",handler="/new/*filepath"} 6
prometheus_http_requests_total{code="200",handler="/service-discovery"} 2
prometheus_http_requests_total{code="200",handler="/static/*filepath"} 54
prometheus_http_requests_total{code="200",handler="/status"} 5
prometheus_http_requests_total{code="200",handler="/targets"} 11
prometheus_http_requests_total{code="302",handler="/"} 20
prometheus_http_requests_total{code="400",handler="/api/v1/query"} 1
prometheus_http_requests_total{code="400",handler="/api/v1/query_range"} 4

We would like here get the number of total requests on handler '/api/v1/query_range', first we are going to collect all the mertics named "prometheus_http_requests_total":

/usr/lib/centreon/plugins/centreon_monitoring_openmetrics.pl \
 --plugin apps::monitoring::openmetrics::plugin \
 --custommode web \
 --mode scrape-metrics \
 --hostname='prometheus.int.centreon.com' \
 --port='9090' \
 --proto='http' \
 --urlpath='/metrics' \
 --filter-metrics='prometheus_http_requests_total'

Command then returns:

OK: All metrics are ok | 'prometheus_http_requests_total'=1;;;; 'prometheus_http_requests_total'=24;;;; 'prometheus_http_requests_total'=3;;;; 'prometheus_http_requests_total'=5454;;;; 'prometheus_http_requests_total'=3390;;;; 'prometheus_http_requests_total'=3;;;; 'prometheus_http_requests_total'=2;;;; 'prometheus_http_requests_total'=2;;;; 'prometheus_http_requests_total'=17;;;; 'prometheus_http_requests_total'=155790;;;; 'prometheus_http_requests_total'=6;;;; 'prometheus_http_requests_total'=2;;;; 'prometheus_http_requests_total'=54;;;; 'prometheus_http_requests_total'=5;;;; 'prometheus_http_requests_total'=11;;;; 'prometheus_http_requests_total'=20;;;; 'prometheus_http_requests_total'=1;;;; 'prometheus_http_requests_total'=4;;;;

We used the options: custommode web (--custommode web) to request the API prometheus using HTTP or HTTPS. Then we configured all the informations about the connections: the hostname (--hostname='prometheus.int.centreon/com'), the port (--port='9090'), the protocol (--proto='http') and the path of the URL where the metrics are accessible (--urlpath='/metrics'). Then we fetched only the metrics 'prometheus_http_requests_totol' (--filter-metrics='prometheus_http_requests_total').

Now we would like filter the requests on URL '/api/v1/query_range', we are going to use the option --instance to have the possibility to filter on a field and then use the option --filter-instance to filter on the value of this fields:

/usr/lib/centreon/plugins/centreon_monitoring_openmetrics.pl \
 --plugin apps::monitoring::openmetrics::plugin \
 --custommode web \
 --mode scrape-metrics \
 --hostname='prometheus.int.centreon.com' \
 --port='9090' \
 --proto='http' \
 --urlpath='/metrics' \
 --filter-metrics='prometheus_http_requests_total' \
 --instance='handler' \
 --filter-instance='/api/v1/query_range'

Command then returns:

OK: All metrics are ok | '/api/v1/query_range#prometheus_http_requests_total'=3390;;;; '/api/v1/query_range#prometheus_http_requests_total'=4;;;;

We filtered on the 'handlers' (--instance='hanlder') and so we only get the metrics where the handler is '/api/v1/query_range' (--filter-instance='/api/v1/query_range').

And now to finish, we are going to filter the requests that return a '200' as the code. And we want to have the possibility to get a graph in Centreon and alerting.

/usr/lib/centreon/plugins/centreon_monitoring_openmetrics.pl \
 --plugin apps::monitoring::openmetrics::plugin \
 --custommode web \
 --mode scrape-metrics \
 --hostname='prometheus.int.centreon.com' \
 --port='9090' \
 --proto='http' \
 --urlpath='/metrics' \
 --filter-metrics='prometheus_http_requests_total' \
 --instance='handler' \
 --filter-instance='/api/v1/query_range' \
 --subinstance='code' \
 --filter-subinstance='200' \
 --warning='5000' \
 --critical='10000'

Command then returns:

OK: All metrics are ok | '/api/v1/query_range~200#prometheus_http_requests_total'=3390;;;;

Here, we added the option --subinstance='code' to have a second filter on the code value, and we filtered on the '200' code value with the opiton --filter-subinstance='200'. For the thresholds, the options --warning='5000' --critical='10000' are set to trigger a WARNING alert if the metric's value is granter than 5000 and a CRITICAL alert if the value is granter than 10000.

All the options of the different modes can be display using the help (--help)

Add the metrics to our monitoring​

We are going to configure the host Prometheus by using the following parameters :

ParameterValue
NAMEprometheus.int.centreon.com
ALIASinternal prometheus
IPADDRESS/FQDNprometheus.int.centreon.com
TEMPLATEApp-Monitoring-Openmetrics-Web-custom
OPENMETRICSPORT80
OPENMETRICSPROTOhttp
OPENMETRICSURLPATH/metrics

image

We can edit the service "Scrape-Metrics" which have been created automatically by using the template "App-Monitoring-Openmetrics-Web-custom" to configure the macros as the same in CLI :

ParameterValue
FILTERMETRICSprometheus_http_requests_total
WARNING5000
CRITICAL10000
INSTANCEhandler
FILTERINSTANCE/api/v1/query_range
SUBINSTANCEcode
FILTERSUBINTACE200
EXTRAOPTIONS--verbose

image

And we rename the service by "Query-Api-Number". This service could be duplicate and modificate to collect other Openmetrics and get alerting/reporting:

image