[PR] UI: Give a Dag level's header its own loading and error states [airflow]

[bbovenzi (via GitHub) via commits]

bbovenzi opened a new pull request, #76391:
URL: https://github.com/apache/airflow/pull/76391

   A run or task instance that failed to load reported it upward to the Dag 
status
   chip, which speaks for the Dag as a whole. A mistyped run id read as a 
problem
   with the Dag, labelled "1 Error" beside its import errors and warnings, with 
the
   reason a click away in a modal. The level that failed is where the failure 
means
   something, so it says so there.
   
   A 404 is shown the same way rather than replacing the panel, because the 
levels
   do not agree on what one means. A task removed from the Dag code 404s because
   the serialized Dag no longer lists it, while its instances and their logs are
   still in the database and still listed by its own tab — so the tab bar stays.
   
   That left the lists under a level repeating what its header had just said, 
since
   the same 404 fails their queries too. A list scoped to a level now leaves a 
404
   to that header and falls back to the empty state it already had. It still
   reports anything else, which is its own failure and nowhere else reported, 
and
   the unscoped pages are unchanged.
   
   The header was also absent until its data arrived. The panel is a fixed 
height
   column whose tab content takes the remaining space, so the card appearing 
took
   that height away and moved everything below it. A skeleton built from the 
card
   itself holds the place: the card is sized by its content, so nothing with a
   hardcoded height would match it.
   
   Which state to show is read from the data, then the failure, then loading. A
   query that keeps failing under auto-refresh reports an error only between its
   attempts and looks pending again while one runs, so deciding in any other 
order
   flips the header between the two every few seconds.
   
   `ErrorAlert` gained an opt-in `closable`, which 
`src/system-components/Alert.tsx`
   already supported but it never forwarded. It is off by default, so its other 
call
   sites are unchanged, and a dismissal is cleared by the next failure.
   
   The mapped and grouped levels keep their current behaviour. Their summaries 
come
   from a stream that reports neither loading nor failure, so there is nothing 
yet
   to show; giving that hook a loading signal is its own change.
   
   ##### Before / after
   
   | | Before | After |
   |---|---|---|
   | **Loading** | <img 
src="https://raw.githubusercontent.com/astronomer/airflow/7dc612de39730bafcd9bc350d206ee294f140f81/before-loading.png";
 width="420"> | <img 
src="https://raw.githubusercontent.com/astronomer/airflow/7dc612de39730bafcd9bc350d206ee294f140f81/after-loading.png";
 width="420"> |
   | **404, missing run** | <img 
src="https://raw.githubusercontent.com/astronomer/airflow/7dc612de39730bafcd9bc350d206ee294f140f81/before-error.png";
 width="420"> | <img 
src="https://raw.githubusercontent.com/astronomer/airflow/7dc612de39730bafcd9bc350d206ee294f140f81/after-error.png";
 width="420"> |
   | **500, run fetch fails** | <img 
src="https://raw.githubusercontent.com/astronomer/airflow/7dc612de39730bafcd9bc350d206ee294f140f81/before-server-error.png";
 width="420"> | <img 
src="https://raw.githubusercontent.com/astronomer/airflow/7dc612de39730bafcd9bc350d206ee294f140f81/after-server-error.png";
 width="420"> |
   
   Loading: the card is absent, so the tabs and everything under them sit about 
a
   hundred pixels high and drop when the run arrives. After, the skeleton holds 
that
   space.
   
   404: no card at all, the reason reaching only the small Dag chip beside the
   breadcrumb, and the same 404 repeated above the table's empty state. After, 
it is
   stated once, on the run's own card, and the chip stays quiet.
   
   500: nothing at all while the fetch fails. After, the header reports it 
while the
   task instance list below, which did not fail, still shows its rows.
   
   ---
   
   ##### Was generative AI tooling used to co-author this PR?
   
   - [X] Yes — Claude Code (Opus 5)
   
   Generated-by: Claude Code (Opus 5) following [the 
guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions)
   
   🤖 Generated with [Claude Code](https://claude.com/claude-code)
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscr...@airflow.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org