fix: retry transient read timeouts when polling dbt Cloud run status

[jymaire]

Problem

io.kestra.plugin.dbt.cloud.CheckStatus (and TriggerRun with wait: true) polls the dbt Cloud run status through AbstractDbtCloud.request, whose retry policy only retries HTTP 502/503/504. A transient read timeout during a status poll is surfaced by the core HTTP client as RuntimeException(java.net.SocketTimeoutException), so it is not retried, propagates out of the polling loop, and fails the whole task — even though the dbt Cloud run is still healthy and goes on to complete:

ERROR java.net.SocketTimeoutException: Read timed out
    at io.kestra.core.http.client.HttpClient.request(HttpClient.java:294)
    at io.kestra.plugin.dbt.cloud.AbstractDbtCloud.lambda$request$1(AbstractDbtCloud.java:94)
    at io.kestra.plugin.dbt.cloud.CheckStatus.fetchRunResponse(CheckStatus.java:214)
    at io.kestra.plugin.dbt.cloud.CheckStatus.run(CheckStatus.java:118)
    at io.kestra.plugin.dbt.cloud.TriggerRun.run(TriggerRun.java:208)

The very next attempt of the same run completed 10 total | 10 success, confirming there was no real dbt error — just a transient network read timeout on one poll.

Root cause

The retry predicate in AbstractDbtCloud.request only matched HttpClientResponseException with status 502/503/504:

(res, throwable) -> throwable instanceof HttpClientResponseException ex &&
    (code == 502 || code == 503 || code == 504)

A read timeout is not an HttpClientResponseException, so it never matched, was not retried, and failed the task.

Fix

Broaden the retry predicate (extracted into isRetriableTransientError) to also treat as transient:

• connection-level failures — HttpClientRequestException (socket / SSL handshake);
• read/connect timeouts — SocketTimeoutException, including when wrapped as the cause.

Genuine client errors (4xx) still fail fast — no change there. Reuses the existing maxRetries / initialDelayMs config; no new properties. Benefits every request path (CheckStatus polling, downloadArtifacts, TriggerRun).

Tests

CheckStatusRetryTest:

• shouldRetryOnReadTimeoutAndEventuallySucceed — reproduces the reported failure (RuntimeException(SocketTimeoutException) then 200) and asserts retry + success. Verified red without the fix (SocketTimeoutException: Read timed out), green with it.
• shouldNotRetryOnClientError — asserts a 404 fails fast (single attempt, no retry).

Full io.kestra.plugin.dbt.cloud package: 14/14 green.

Closes #295

🤖 Generated with Claude Code

[github-actions]

📦 Artifacts

Name	Size	Updated	Expiration
	212.19 KB	Jul 6, 26, 3:42:13 PM UTC	Jul 13, 26, 3:42:12 PM UTC

🧪 Java Unit Tests

	Tests	Passed ✅	Skipped	Failed	Time ⏱
Java Tests Report	40 ran	40 ✅	0 ⚠️	0 ❌	4m 2s 331ms

🔁 Unreleased Commits

✅ No unreleased commits found.

[github-actions]

Tests report quick summary:

success ✅ > tests: 40, success: 40, skipped: 0, failed: 0

unfold for details

Project	Status	Success	Skipped	Failed
plugin-dbt	success ✅	40	0	0