The general observation seemed to be that it makes sense to reduce the retention size to about 80-85% of the total allocated storage, to allow for time series entries to be culled prior to running out of space.
I'll start the PR for this but wanted to link it to this issue in case others would like to weigh in with more specific guidance.
Based on a conversation in the Prometheus Slack: https://cloud-native.slack.com/archives/C167KFM6C/p1720815634650289
The general observation seemed to be that it makes sense to reduce the retention size to about 80-85% of the total allocated storage, to allow for time series entries to be culled prior to running out of space.
I'll start the PR for this but wanted to link it to this issue in case others would like to weigh in with more specific guidance.