Is there a possibility to limit the writes/s for different output plugins?
Particularly, this would be relevant when emptying the metric_buffer after an output was unavailable. The current behaviour leads to a massive write peak as soon as the output is availabe again and therefore overloading the subsequent processing tasks.
Data is flushed to outputs in very specific cases. Generally, I would suggest someone increase their batch size or interval to decrease how often data is pushed to output.
However, you mentioned that the output was unavailable. Unfortunately, in Telegraf there is currently no way to break up a pending buffer once the output is restored.
One option that some use is when they have a noisy network or potential network disconnects/loss between telegraf and a source is to put a message queuing system (e.g. rabbitmq, kafka) between their data collection and telegraf, so telegraf only grabs so many items at once. Then telegraf is only grabbing N number of data points when the source is available and when it is not available nothing is getting recorded.
