-
Notifications
You must be signed in to change notification settings - Fork 5.8k
Max ttl for gauges in statsd input plugin #8348
Copy link
Copy link
Closed
Labels
area/statsdfeature requestRequests for new plugin and for new features to existing pluginsRequests for new plugin and for new features to existing pluginsgood first issueThis is a smaller issue suited for getting started in Telegraf, Golang, and contributing to OSS.This is a smaller issue suited for getting started in Telegraf, Golang, and contributing to OSS.help wantedRequest for community participation, code, contributionRequest for community participation, code, contribution
Metadata
Metadata
Assignees
Labels
area/statsdfeature requestRequests for new plugin and for new features to existing pluginsRequests for new plugin and for new features to existing pluginsgood first issueThis is a smaller issue suited for getting started in Telegraf, Golang, and contributing to OSS.This is a smaller issue suited for getting started in Telegraf, Golang, and contributing to OSS.help wantedRequest for community participation, code, contributionRequest for community participation, code, contribution
Feature Request
Proposal:
Allow for setting "gaugesMaxTTL" option in the
statsdinput plugin. Essentially, this from the core statsd repo:https://github.com/statsd/statsd/pull/599/files
Current behavior:
If
deleteGaugesis set to "False" then once a gauge metric is reported it is reported for the life of thetelegrafserver. There is no way to set a TTL value where if the gauge metric is not received at the ingress endpoint after a certain time it is reset to 0.Desired behavior:
An optional
gaugesMaxTTLoption is available for thestatsdinput -- this has exactly the same effect as thedeleteGaugesoption, except that it only executes that logic after no input has been received for the value specified ingaugesMaxTTL.Use case:
Telegraf server
Aon hostXis taking input from serviceBand serviceC. Both of these services are reporting metrics via a gauge with "deleteGauges" set toFalse-- let's assume this is required w/o getting distracted by details. :-)Now, service
Cis turned down by some load-managing automation. However, telegrafAwill still report the last gauge metrics to Influx until restarted, with no way of ever reflecting the fact that serviceCis permanently gone.