Hi,
In my current setup, children are streaming metrics to a parent without any child-local storage at all. This causes gaps when the parent restarts, presumably because the child failed to stream and throws away the metrics.
Is there any way to avoid this without enabling local storage on the child nodes, or is local storage there a requirement to avoid this issue?
The “expected behavior” for me would be that the children have some kind of in-memory buffer of a certain size to make them tolerant to short disconnects (minutes, not hours), but perhaps this already exists but only for nodes where disk storage is enabled?
Thanks in advance.