end timestamp must not be before or equal to start time (Loki)

[loganmc10]

What happened:
Sometimes (not always) my time series graph will fail to load with the error shown in the screenshot.

What you expected to happen:
Graph should show data

How to reproduce it (as minimally and precisely as possible):
Not exactly sure, this seems to be new.

Anything else we need to know?:
This is using the new time series panel

Environment:

• Grafana version: 7.4.5
• Data source type & version: Loki, Grafana Cloud
• OS Grafana is installed on: Grafana Cloud
• User OS & Browser: Windows, Chrome 89
• Grafana plugins: N/A
• Others:

[loganmc10]

When it works (if I refresh a few times), it looks like this:

[ivanahuckova]

Hello @loganmc10m, what Loki version are you using? Could be related to #3286. Also I remember that @dannykopping had the same issue. Have you found a reason for these errors?

[loganmc10]

I'm using Grafana Cloud so I'm not exactly sure what version of Loki it is, some recent version though I assume.

I haven't figured out a reliable way to reproduce it, but it's still happening. Refreshing the dashboard a few times makes it go away

[ivanahuckova]

@cyriltovena I am passing this to Loki repo as I think this is not related to Grafana. It is possible duplicate of #3286, but I would rather leave that to you to determine.

Also, couple weeks ago we were debugging the same issue with Danny. We were getting error for queries that went past 7 days. What was interesting - the metrics queries for that time range failed, but log queries for the exactly same time range worked.

Also, if you think this is not Loki issue, feel free to transfer back and I'll try to look at it again.

[darioef]

I'm getting the same error. It's not related to which panel you're using. It happened to me with Stats, Diagram, Table, etc.

I found that it doesn't happen if you avoid Query Frontend and do the query directly to the Querier but I can't find which setting is the resposible of this behaviour.

"Max Interval" in Grafana and "split_queries_by_interval" in Loki have something to do with this but I can't find how.

Loki version: 2.2.0 running on Docker Swarm (own deployment).

[cyriltovena]

This has been fixed by #3540 thanks for the report.