When scrape interval is used, metrics are not available outside the scrape window

[redelang]

When using scrape interval on global level you can also define scrape interval per metric. Metrics which have an individual scrape interval configured in the configuration (toml) file, are only available at the moment it's actually scraped from the database. On the next global scrape interval the metric is no longer available. My expectation is that the metric would still be available, but with the old value.

e.g. global scrape interval is 2 minutes and an individual metric has a scrape interval of 1 hour configured. In this case that metric is only available for 2 minutes per hour.

[anders-swanson]

@redelang thanks for reporting this, we'll take a look.

[anders-swanson]

I believe what's happening is the go-prometheus SDK doesn't "remember" metrics - it only serves the last snapshot of metrics. There's a way to refactor the exporter to keep the whole snapshot in memory, which seems the way to go for this kind of issue.