In job backfill UI, make it clear that backfill policies are not respected

[sryza]

Users have (reasonably) expressed confusion about the fact that asset job backfills don't respect backfill policies.

The root of this is that job partitions are defined in terms of runs: i.e. when you launch a job backfill, you're telling Dagster that you want a run for every partition covered by the backfill. The Partitions Tab on the job page is based around run-based accounting. So supporting backfill policies for asset job backfills would require radical changes to the job partitions data model.

We may be able to eventually make this radical change, but in the mean time, we can still help users avoid a pit of failure. When users try to launch a job backfill for assets that have backfill policies other than the default one, we could do one of:

• Show a big banner warning them that they might not get the behavior they expect
• Redirect or nudge them towards launching a direct asset backfill on the assets covered by the job

Related issue: #11962

What we've heard

• https://dagsterlabs.slack.com/archives/C05NZTUSMPU/p1698929833987769?thread_ts=1698256384.379429&cid=C05NZTUSMPU

[qgab-flowdesk]

EDIT: Since 1.5 the partitions status is not specified anymore in my pipelines. A daily partition with backfill_policy = BackFillPolicy.single_run() would executed, succeeed, and yet be marked as missing

Having the same issue here:

Not sure to understand how the single run feature that was available before is clashing with the current data model? The second option of redirecting toward the old single run option would be a life saver for dbt users.

[jPinhao-Rover]

This is a pretty confusing limitation. Is there a way to prevent a job from triggering if it is a backfill? We want to control how we run backfills for large models, and if someone triggers 4 years worth of partitions for a daily partitioned job without noticing, it's going to be pretty damning :s

We can limit concurrency using tags, but is there a way to identify a job was triggered for a backfill and actually prevent it from running?

[qgab-flowdesk]

EDIT: Since 1.5 the partitions status is not specified anymore in my pipelines. A daily partition with backfill_policy = BackFillPolicy.single_run() would executed, succeeed, and yet be marked as missing

Having the same issue here:

Not sure to understand how the single run feature that was available before is clashing with the current data model? The second option of redirecting toward the old single run option would be a life saver for dbt users.

Just wanted to say that this very problematic issues has been fixed in this commit. Thank you guys!