Hi Tony,
Yes, for those fresh log files, they are queriable right away, I updated this part as well in my reply before your telnet suggestion.
Then I would say it’s working as expected.
If you send a log with timestamp in the past, it would still need to sit on the ingester until the max idle time is reached before it’s written into the storage.
Yes, that’s correct, but if they are in the ingester, why querier cannot query it directly from there till it is de-staged to backend storage?
That is expected, right?
Let’s say query_ingesters_within is set to 3h, and you send a log into Loki that’s 4 hours in the past. And let’s say you query for the past 6 hours:
For hour -o to -3, querier goes to ingester, which does not find log that’s 4 hours old.
For hour -3 to -6, querier goes to storage, which does not find the log that’s still waiting for max idle on ingester.
Thanks Tony.
So the value of query_ingesters_within is the actual timestamp of the log files (such as
2024-03-13 23:53:51,980+0800
) but not the time that I ingested the log files, is that right?
I… guess?
But it’s same for any query, right? Imagine if you have a log in loki with a timestamp from 24 hours ago, and you send a query for the past 23 hours, you won’t see that log. Same idea here, if you have a lot in ingester with a 4 hours timestamp and you send a query to ingester for the past 3 hours you won’t see that log.
I think you are thinking about this a bit too hard 
All right, if that is the case then the behavior makes sense now.
Thanks for your patience and prompt reply all the way long.