ℹ️ SignalFx was acquired by Splunk in October 2019. See Splunk SignalFx for more information.
A collectd plugin for aggregating and reporting metrics made available by the kong-plugin-signalfx status endpoint. It is intended to be run within the collectd-python embedded interpreter.
collectd-kong
and its dependencies can be installed within a system Python's site-packages, but to
reduce the potential for dependency conflicts and to account for the collectd embedded interpreter's
packaging limitations, using pip's --target
option and a standalone package directory is recommended:
git clone https://github.com/signalfx/collectd-kong.git
mkdir /opt/collectd-plugin
pip install --target /opt/collectd-plugin collectd-kong/
pip install --target /opt/collectd-plugin/kong -r collectd-kong/requirements.txt
cp collectd-kong/kong_plugin.py /opt/collectd-plugin/kong_plugin.py
Kong Admin API connectivity settings and desired aggregation and reporting configurations are specified similar to other Python collectd plugins
LoadPlugin python
<Plugin python>
ModulePath "/opt/collectd-plugin/kong"
ModulePath "/opt/collectd-plugin"
LogTraces true
Import "kong_plugin"
<Module kong_plugin>
URL "https://my_kong_server:8443/signalfx"
AuthHeader "Authorization" "Basic YWRtaW46cGFzc3dvcmQ="
CABundle "/path/ca_bundle/"
VerifyCerts true
ClientCert "/path/client.cert"
ClientCertKey "/path/client.key"
Metric "request_latency" false
Metric "request_size" false
Metric "response_size" true
Metric "upstream_latency" true
ReportHTTPMethods true
ReportStatusCodes true
ReportStatusCodeGroups false
ReportRouteIDs true
ExtraDimension "my_custom_indentifier" "012abcABC"
Interval 20
Verbose true
</Module>
</Plugin>
For any Kong object that has had kong-plugin-signalfx
registered via the Admin API, metrics will be made
accessible to the /signalfx
status endpoint.
Metric | Metric directive first value | Description | Default reporting value |
---|---|---|---|
counter.kong.connections.accepted |
"connections_accepted" |
Total number of all accepted connections. | false |
counter.kong.connections.handled |
"connections_handled" |
Total number of all handled connections (accounting for resource limits). | false |
counter.kong.kong.latency |
"kong_latency" |
Time spent in Kong request handling and balancer (ms). | false |
counter.kong.requests.count |
"total_requests" |
Total number of all requests made to Kong API and proxy server. | true |
counter.kong.requests.latency |
"request_latency" |
Time elapsed between the first bytes being read from each client request and the log writes after the last bytes were sent to the clients (ms). | true |
counter.kong.requests.size |
"request_size" |
Total bytes received/proxied from client requests. | true |
counter.kong.responses.count |
"response_count" |
Total number of responses provided to clients. | true |
counter.kong.responses.size |
"response_size" |
Total bytes sent/proxied to clients. | true |
counter.kong.upstream.latency |
"upstream_latency" |
Time spent waiting for upstream response (ms). | true |
gauge.kong.connections.active |
"connections_active" |
The current number of active client connections (includes waiting). | false |
gauge.kong.connections.reading |
"connections_reading" |
The current number of connections where nginx is reading the request header. | false |
gauge.kong.connections.waiting |
"connections_waiting" |
The current number of idle client connections waiting for a request. | false |
gauge.kong.connections.writing |
"connections_writing" |
The current number of connections where nginx is writing the response back to the client. | false |
gauge.kong.database.reachable |
"database_reachable" |
kong.dao:db.reachable() at time of metric query | false |
Aggregating these Kong object-level metrics as collectd metric values is done by defining desired context
groups. The reported request/response context group granularity for these metrics' time series is provided via
dimensions,
which will be included for each target aggregation group whose membership is determined by configurable Report*
boolean directives:
Directive | SFx dimension | Description | Default reporting value |
---|---|---|---|
ReportAPIIDs |
api_id |
The UUID assigned to each API object upon creation | true |
ReportAPINames |
api_name |
The optional, user-created name assigned to API objects (recommended) | true |
ReportServiceIDs |
service_id |
The UUID assigned to each Service object upon creation | true |
ReportServiceNames |
service_name |
The optional, user-created name assigned to Service objects (recommended) | true |
ReportRouteIDs |
route_id |
The UUID assigned to each Route object upon creation (recommended) | true |
ReportHTTPMethods |
http_method |
The HTTP method of each request made to the Kong proxy | true |
ReportStatusCodeGroups |
status_code |
The HTTP status code group (e.g. 4xx) for each proxy server fielded request (recommended) | true |
ReportStatusCodes |
status_code |
The HTTP status code for each fielded request | false |
With the exception of ReportStatusCodeGroups
, each of the Report*
directives has an associated white and blacklist
for fine tuning the desired dimensions to include with each datapoint by an arbitrary number of unix
filename patterns. By utilizing these lists kong-plugin-signalfx
can be enabled globally, but only a near minimum of
time series will be created for closely monitoring a select number of object and response contexts without losing
the ability to monitor the system as a whole.
Directive | Description |
---|---|
APIIDs |
The pattern(s) of API IDs to add as dimensions for applicable datapoints |
APIIDsBlacklist |
The pattern(s) of API IDs to not report as dimensions for applicable datapoints |
APINames |
The pattern(s) of API Names to add as dimensions for applicable datapoints |
APINamesBlacklist |
The pattern(s) of API Names to not report as dimensions for applicable datapoints |
ServiceIDs |
The pattern(s) of Service IDs to add as dimensions for applicable datapoints |
ServiceIDsBlacklist |
The pattern(s) of Service IDs to not report as dimensions for applicable datapoints |
ServiceNames |
The pattern(s) of Service Names to add as dimensions for applicable datapoints |
ServiceNamesBlacklist |
The pattern(s) of Service Names to not report as dimensions for applicable datapoints |
RouteIDs |
The pattern(s) of Route IDs to add as dimensions for applicable datapoints |
RouteIDsBlacklist |
The pattern(s) of Route IDs to not report as dimensions for applicable datapoints |
HTTPMethods |
The pattern(s) of HTTP methods to add as dimensions for applicable datapoints |
HTTPMethodsBlacklist |
The pattern(s) of HTTP methods to not report as dimensions for applicable datapoints |
StatusCodes |
The pattern(s) of status codes to add as dimensions for applicable datapoints |
StatusCodesBlacklist |
The pattern(s) of status codes to not report as dimensions for applicable datapoints |
Directive values are case sensitive and blacklists will always take precedence over matching
whitelist values. Please note that ReportStatusCodeGroups
and ReportStatusCodes
may not be specified at the same
time, but whitelisted HTTP status codes can be used in tandem with the ReportStatusCodeGroups
directive. This allows
reporting metics associated with specific codes, while not including them with the Nxx
group aggregation.
As an example, here is a collectd configuration file for a globally monitored Kong system and aggregation by desired routes, HTTP methods, and status codes:
LoadPlugin python
<Plugin python>
ModulePath "/opt/collectd/kong"
LogTraces true
Import "plugin"
<Module plugin>
URL "http://my_kong_server:8001/signalfx"
ReportHTTPMethods false
HTTPMethods "PATCH" "DELETE"
ReportStatusCodes false
ReportStatusCodeGroups true
StatusCodes 419 500
ReportRouteIDs false
RouteIDs "689ab4ac-eef0-418e-b4c1-e32f1fa80032"
RouteIDs "ca97d13f-9ab1-49fe-8184-092d66263598"
</Module>
</Plugin>
In this case no http_method
dimensions will be reported other than "PATCH"
and "DELETE"
and no non-grouped
status_code
dimensions will be reported other than 419
and 500
. Similarly, no route_id
dimensions will be
provided for any Route other than the two desired. All other aggregated metric reporting will follow the default
directive values.
There are also optional configuration directives for general behavior and Kong Admin API connectivity:
Directive | Description | Default |
---|---|---|
URL |
The URL to reach the kong-plugin-signalfx status metrics. | "http://localhost:8001/signalfx" |
AuthHeader |
The name and value of a header to be passed with GETs to the URL | None |
VerifyCerts |
Whether to verify the ssl certificates for HTTPS requests to the URL | true |
CABundle |
Path to a CA_BUNDLE file or directory with certificates of trusted CAs when VerifyCerts is true |
None |
ClientCert |
Client side certificate to use for HTTPS requests to the URL | None |
ClientCertKey |
Separate client side certificate key if not included with cert file | None |
Interval |
How often, in seconds, Kong metrics are obtained. | None, inherits collectd setting |
Verbose |
Whether to emit lower level metric collection and processing statements to the LogFile plugin | false |