Feature Request
Make it possible to start checking on intervals that are NOT round_interval - most useful for measuring remote response.
Proposal:
Implement a displacement that will be added to round_interval.
Current behavior:
interval = "10s"
collection_jitter = "0s"
round_interval = true
causes collection to start at :00, :10, :20 etc
Desired behavior:
interval = "10s"
displacement = "5s"
collection_jitter = "0s"
round_interval = true
causes collection to start at :05, :15, :25 etc
Use case:
To provide the measurements for SLA reporting, we have several servers used for remote monitoring.
To simplify, assume a customers web-service. To monitor it we configure telegraf on two different servers.
Both of these telegraf polls the web-service. The reason for having two servers is that we always have monitoring of the service - even when after patching the telegraf-servers (obviously assuming that we do not patch them at the same time), or having any other downtime on one of the telegraf-servers.
In normal situations (i.e. when both telegraf-servers are up), we would like to poll the web-service (almost) equidistantly. This cannot be achieve with todays configuration. With a displacement that would be easy and this would also reduce the time to detect that a service is down.
Feature Request
Make it possible to start checking on intervals that are NOT round_interval - most useful for measuring remote response.
Proposal:
Implement a displacement that will be added to round_interval.
Current behavior:
causes collection to start at
:00,:10,:20etcDesired behavior:
causes collection to start at
:05,:15,:25etcUse case:
To provide the measurements for SLA reporting, we have several servers used for remote monitoring.
To simplify, assume a customers web-service. To monitor it we configure telegraf on two different servers.
Both of these telegraf polls the web-service. The reason for having two servers is that we always have monitoring of the service - even when after patching the telegraf-servers (obviously assuming that we do not patch them at the same time), or having any other downtime on one of the telegraf-servers.
In normal situations (i.e. when both telegraf-servers are up), we would like to poll the web-service (almost) equidistantly. This cannot be achieve with todays configuration. With a
displacementthat would be easy and this would also reduce the time to detect that a service is down.