This message was deleted.

Generally the issue is historical loading cache size. https://druid.apache.org/docs/latest/configuration/#storing-segments To check why the segments are not loaded : 1. Restart the historical. 2. Check the coordinator logs. 3. Check the historical logs. Also druid 27.0 has a ton of updates to the segment loading piece called smart segment loading which is turned on by default. I suggest you try that

We have millions of segments. If we restart the historical, it will take days to get the data available. Do you suggest any other way ? @Karan Kumar

I think in that case you should look at compaction or MSQ replace statements.

You could consider adding a few historicals with a separate tier and then changing the retention rules to that new tier only for the problematic datasource

@Amatya Avadhanula separate tier you mean ? Can you share any of its related docs please ?

@sriramdas sivasai you can use this parameter

druid.segmentCache.lazyLoadOnStart=true

it will bring up the druid but it will load the segments when it is needed by the query