Logs query usage — climbing with no queries I can find?

The short version: my logs query usage has been climbing steadily for about 40 hours and I can’t figure out what’s actually running the queries. grafanacloud_logs_instance_query_overage has been rising at a very consistent ~7.5/hr since roughly June 20, 22:00 UTC, and it’s now around 312 — which is strange, because I only ingest about 59 MiB of logs total. So it’s almost entirely query volume, not data.

What’s got me stuck is that I can’t find the queries. In the usage-insights request log ({instance_id=“1611752”} |= “request timings”), the only thing hitting the instance since about June 22, 03:47 UTC is my three alert rules doing quick instant queries every 60s. (Before that there were some query_range queries under my own username, but they stop in the log at 03:47 — and the usage kept right on climbing afterward anyway, so they don’t seem to explain it.) Meanwhile grafanacloud_logs_instance_query_bytes:rate5m on the querier is still sitting around 700–800M right now. So a couple of metrics say something’s scanning a lot, but I can’t find any matching query in the log — and it’s been climbing all night while my laptop was asleep.

I did look into the Loki Query Fair Usage dashboard, but nothing really jumped out. The top “bytes-by-source” result was “grafana-alerts”, but everything was showing in the 10s of MiB, so that doesn’t seem like it could be the thing.

I’ve gone through everything I can think of on my end — revoked all my sessions, checked my access-policy tokens (the only one is write-only, so it can’t query), no service accounts of mine, no dashboards or recording rules pointing at the logs — and I’m honestly out of ideas. Any ideas for how to continue to triage this would be very helpful!

Run your alert queries in Explore and check This query will process approximately info. Multiply it based on your alert period and estimate log query bytes used for your selected period.

My guess is that your alert queries are causing that loq query usage.

If I remove the outer functions and just put the query in the explorer, it says the query will process 44 KiB. And this query runs every 60 seconds, so that’s not plausibly the culprit.

44kb * 3 queries * 60 * 24 * 30 /1024 = 5.5GB of log reading per month for alerting
But you are claiming “about 59 MiB of logs total” (you are really missing time units in your statements, so I guess it is per month)
So you are writing 59Mib per month and reading 5.5GiB per month = you are close to that fair usage, so you very likely overusing your fair Grafana Cloud limits.

I would say go and ask your Grafana Cloud support for explanation. They have proper access to your instance stats.