The variable $__interval is templated to an incorrect value during annotation query

[Haleygo]

To reproduce:
Adding an annotation with following query:
sum(changes(vm_app_start_timestamp{job=~"$job", instance=~"$instance"}[$__interval])) by(job)
and execute this query over a big time range like 24h, see this example on playground.
VM-datasource sets the $__interval to 60s, but the actual step for the returned results is 15m(probably limited by returned data points), resulting in missed restart events.

Expected result:
$__interval should be templated as the actual step arg in the query request to datasource as prometheus datasource, see this example.

[dmitryk-dk]

HI @Haleygo ! The issue was fixed in the new v0.10.3 release. Close as solved. Fill free to reopen if you find any problem