| Warning |
|---|
This document is not ready for publication. Remove this notification and remove viewer restrictions when ready. |
...
The ID Service stores metrics data to an InfluxDB service with a default retention policy of 24 hours. This may not be long enough time for your use cases. This document will show with examples how to store metrics data for a longer period of time, and how to configure Grafana to display it.
| Table of Contents |
|---|
The theory
To store measurements for a longer time, it must be stored with a Retention Policy of desired length. It doesn’t make sense to store ALL data for the specified time: When Grafana shows values from a month in a graph, it physically cannot display data at 50 millisecond accuracy. Having a lot of data at very short intervals takes a lot of processing, bandwidth and disk space. So you must think about downsampling the data as well.
...
Let’s create a CQ that collects everything usable from default mgmt_api_request data. We can find the tag and field names with commands like SELECT * FROM mgmt_api_request LIMIT 1.
...
| breakoutMode | full-width |
|---|
...
CREATE
...
CONTINUOUS
...
QUERY
...
"mgmt_api_requests_5min_for_1month"
...
ON
...
"oneportal"
...
BEGIN
...
SELECT
...
count(auth_header_exists)
...
AS
...
"count_requests",
...
sum("request_duration")
...
AS
...
"sum_request_duration",
...
mean("request_duration")
...
AS
...
"mean_request_duration"
...
INTO
...
"one_month"."mgmt_api_requests"
...
FROM
...
"mgmt_api_request"
...
GROUP BY
...
time(5m),
...
"path",
...
"method",
...
"api_client_id",
...
"oauth2_client_id"
...
END
We could have created a CQ that collects only the number of requests and groups them by path, but in anticipation of other needs, we collect more data and more tags.
...
In FROM use the
one_monthRP and themgmt_api_requestsname,change the SELECT parameters to use field
count_requestsand the function tosum()since we now need to add the counts up.
...
The equivalent query is: SELECT sum("count_requests") FROM "one_month"."mgmt_api_requests" WHERE $timeFilter GROUP BY time($__interval), "path" fill(0).
The graph now shows the expected correct values. You should change the Min interval to 5m to match the CQ, and to visualise that the data covers a 5 minute period.
...
More CQ examples
Number of Management API requests that ended if error code (>= 400):
CREATE CONTINUOUS QUERY mgmt_api_request_errors_5min_for_1month ON oneportal BEGIN SELECT count(auth_header_exists) AS count_requests INTO oneportal.one_month.mgmt_api_request_errors FROM oneportal.one_day.mgmt_api_request WHERE response_code >= 400 GROUP BY time(5m), api_client_id, oauth2_client_id, path END
Number of Management API requests that ended in successful code (200 to 399):
CREATE CONTINUOUS QUERY mgmt_api_request_successes_5min_for_1month ON oneportal BEGIN SELECT count(auth_header_exists) AS count_requests INTO oneportal.one_month.mgmt_api_request_successes FROM oneportal.one_day.mgmt_api_request WHERE response_code >= 200 AND response_code < 400 GROUP BY time(5m), api_client_id, oauth2_client_id, path END
Number of OpenID API requests, similar to number of Management API requests in the above example:
CREATE CONTINUOUS QUERY "openid_api_requests_5min_for_1month" ON "oneportal" BEGIN SELECT count(auth_header_exists) AS "count_requests" INTO "one_month"."openid_api_requests" FROM "openid_api_request" GROUP BY time(5m), "path", "method" END
Number of log entries:
CREATE CONTINUOUS QUERY "log_entries_5min_for_1month" ON "oneportal" BEGIN SELECT count("event_id") AS "count_entries" INTO "one_month"."log_entries" FROM "log_entry" GROUP BY time(5m), "source" END