Fix dag duration metric

[jward-bw]

The filter was causing an implicit cartesian join on TaskInstance in the outer query, making this metric appear multiple times for a single dag. I have moved the filter into the sub-query which is already joined with TaskInstance to prevent this cartesian join.

@abhishekray07 I hope you can take a look at this soon.

[jward-bw]

This PR addresses this issue: #24

[mshalak-nix]

+1 for this fix, we have ~28K task instances, and "dag_run_duration" metric gets duplicated 28K times.
@ayush-san @abhishekray07

[drboyer]

Any idea when these changes will be merged and released @ayush-san @abhishekray07?

[ayush-san]

@drboyer I don't have merging rights. Waiting for @abhishekray07 to merge and release these changes

[dusty73]

We're also experiencing performance issues, after a week the metrics endpoint takes something like 20 seconds to respond. When do you think this fix will be merged?

[jward-bw]

The owner of this repo seems to be inactive, so you'd do better just forking and applying this patch yourself.

[azwjp]

FYI, issue #26 is also related to this PR.

[ring-pete]

Hello just bumping this PR to see if any upstream maintainers will be able to merge this soon :)

[jonathonbattista]

For anyone having performance or query response time issues with this exporter, this change will fix those.