ingester: remove unused loki_ingester_sent_chunks metric

[rfratto]

While loki_ingester_sent_chunks was given a value, by the time Prometheus would scrape it, the ingester would have already shut down and would never have a non-zero value.

loki_ingester_received_chunks stores the same value on the receiving end and lives long enough to be picked up by Prometheus.

[cyriltovena]

LGTM should you see if we have to update a dashboard now ?

[rfratto]

This wasn't in any dashboards since it's brand new so nothing should have to be updated.

However, I talked about this with @tomwilkie and he mentioned that the ingester lifecycler is supposed to sleep for long enough so the metric can be scraped. Seems like that's not happening for us; I need to look into why. Closing.

[cyriltovena]

@pracucci I think did this change ?

[rfratto]

It looks like the default wait period is long enough but the server serving the metrics shuts down before anything else so any shutdown metrics would never be scraped. I'll open an issue for this

[pracucci]

@cyriltovena In the PR #784 I've set final_sleep: 0s in the cmd/loki/loki-local-config.yaml. The rationale behind that PR was that - given that's a sample local config file - we should aim for a fast shutdown instead of forcing a 30s (default) sleep on shutdown.

I think the problem described here by @rfratto is different, assuming he's not using the loki-local-config.yaml.

[rfratto]

@pracucci Yeah, this appears to be caused by #819 and isn't related to your change 🙂

[cyriltovena]

thanks !