-
Type: Improvement
-
Status: Resolved
-
Priority: Minor
-
Resolution: Fixed
-
Affects Version/s: 11.1
-
Component/s: Monitoring
-
Epic Link:
-
Upgrade notes:
-
Team:PLATFORM
-
Sprint:nxplatform 11.1.29, nxplatform 11.1.30, nxplatform 11.1.31, nxplatform 11.1.32
-
Story Points:0
Today, Nuxeo publishes its Dropwizzard'smetrics to Datadog using the following addon:
https://github.com/nuxeo/nuxeo/tree/master/addons/nuxeo-datadog-reporter
To build a useful dashboard, metrics need to be manipulated using wildcards, for instance:
# to aggregate the number of errors sumSeries(servers.*.nuxeo.org.apache.logging.log4j.core.Appender.{error,fatal}.count)" # or to render the scheduled works in each workmanager pool: aliasByNode(servers.*.nuxeo.nuxeo.works.total.*.scheduled, 1, 6)
But wildcards are not supported by Datadog the dimension for a metric should be expressed using tagging not using the metric name,
we should rewrite the metric to:
metric: nuxeo.log4j tagging: server: <server-name> loglevel:<pool-name> formula: sum:nuxeo.log4j{$var} by {loglevel} metric: nuxeo.works.scheduled tagging: server: <server-name> workpool:<pool-name> formula: sum:nuxeo.works.scheduled{$var} by {workpool}
Some investigation must be done on the addon to see if the latest [coursera lib|https://github.com/coursera/metrics-datadog] provides a way (maybe using DynamicTagsCallback) to create dynamically tagging based on the metrics name.
Using tagging will reduce drastically the number of metrics in Datadog which can reduce the cost of monitoring.