fix(source-salesloft, source-aws-cloudtrail): bump stale SDM base image pin - #84350
fix(source-salesloft, source-aws-cloudtrail): bump stale SDM base image pin#84350devin-ai-integration[bot] wants to merge 2 commits into
Conversation
Co-Authored-By: bot_apk <apk@cognition.ai>
🤖 Devin AI EngineerI'll be helping with this pull request! Here's what you should know: ✅ I will automatically:
Note: I can only respond to comments from users who have write access to this repository. ⚙️ Control Options:
|
👋 Greetings, Airbyte Team Member!Here are some helpful tips and reminders for your convenience. 💡 Show Tips and TricksPR Slash CommandsAirbyte Maintainers (that's you!) can execute the following slash commands on your PR:
📚 Show Repo GuidanceHelpful Resources
|
|
Note Autopilot progressive rollouts are not enabled for the following modified connector(s):
This is a courtesy heads-up only — it does not block merge or fail any check. |
|
Deploy preview for airbyte-docs ready!
Deployed with vercel-action |
|
Co-Authored-By: bot_apk <apk@cognition.ai>
|
|
CI status: all mechanical checks pass (format, docs build, metadata/changelog validation, connector lint, static analysis, progressive rollout). The connector acceptance tests fail for both connectors, and as far as I can tell the failures are credential-related rather than caused by this bump — but I can't fix them from here, so flagging for a maintainer's call. Show/Hide Evidence
Why this looks pre-existing rather than caused by the base image bump: running That points at stale acceptance-test credentials in GSM for both connectors. Refreshing the GSM acceptance-test secrets for both connectors would be needed to get these green. I have not touched |
What
source-salesloftandsource-aws-cloudtrailwere the only two connectors left in the monorepo pinningconnectorBuildOptions.baseImagetosource-declarative-manifest:5.15.0(365 connectors are on7.24.0). That image has noairbyteuser, while the shareddocker-images/Dockerfile.manifest-only-connectorends withUSER airbyte(added in #64938). Both connectors therefore build fine and then fail at container start:Already-published images are unaffected — they predate the
USER airbytechange — but anything that builds from source is blocked: the pre-release publish workflow, the regression-test harness (fails atSPECwithInternal failure: true), and each connector's own CI image/spec check.This unblocks #84300 and airbytehq/oncall#13289 (the Salesloft OAuth refresh fix), which cannot be published or regression-tested until a source build runs. #84300 also bumps Salesloft; whichever lands second needs to take the next patch version.
Resolves https://github.com/airbytehq/oncall/issues/13301:
Community report: #84349
Requested by
aaronsteersvia the/ai-fixworkflow.How
Base image pin moved to the
7.24.0pin used by the rest of the monorepo, plus a patch bump and changelog row per connector.manifest.yamlandcomponents.pyare untouched in both connectors.Neither connector has
enableProgressiveRollout/rolloutConfiguration, so plain semver with no-rcsuffix. Neither has apyproject.toml— for manifest-only connectors the base image pin is the CDK dependency, so this is a CDK 5 → 7 jump for both.Review guide
airbyte-integrations/connectors/source-salesloft/metadata.yaml— base image5.15.0→7.24.0,1.5.2→1.5.3airbyte-integrations/connectors/source-aws-cloudtrail/metadata.yaml— same base image change,1.1.0→1.1.1(the existing1.0.0breakingChangesentry is untouched)docs/integrations/sources/salesloft.md,docs/integrations/sources/aws-cloudtrail.md— changelog rowsBreaking change evaluation
Assessed as not breaking, so a patch bump rather than a major. Evidence: both connectors were built locally against the
7.24.0pin anddiscoverwas run with the real GSM acceptance-test configs. The catalogs came back unchanged — Salesloft: 30 streams, PKid, incremental cursorupdated_at; CloudTrail:management_events, PKEventId, cursorEventTime. No spec change, no stream added or removed, no state-format change. All declarative component types in both manifests are recognized by CDK 7.24.0, and each manifest's declaredversion:(5.15.0and4.3.2) still loads. Bothcomponents.pymodules import and construct cleanly inside the 7.24.0 image.Test Coverage
No unit test is added. Neither connector has any test infrastructure (no
unit_tests/, nopyproject.toml), and the change is a build/packaging pin with no connector logic to assert on — a unit test would run against the installed CDK, not against the base image the pin selects. Verification was done at the image level instead:docker-images/Dockerfile.manifest-only-connectorwithBASE_IMAGEset to the new pin; both images start and return a validSPEC. The same build against the old5.15.0pin reproduces the exit-125unable to find user airbytefailure.discoversucceeds for both connectors on7.24.0, with the dummy config and with the GSM config.checkfails for both, but identically on5.15.0(run with an overridden container user so the old image can start at all) and on7.24.0, so it is pre-existing and not caused by the bump: Salesloft returnsHTTP 401 Invalid Bearer tokenand CloudTrail returnsUnrecognizedClientException: The security token included in the request is invalid, i.e. the stored acceptance-test credentials are stale in both cases. For CloudTrail the signed request body, endpoint, target, content type, and SigV4 signed-header set were compared between the two CDK versions and match, so its custom SigV4components.pyis not affected by the CDK jump.CI's connector image/spec checks are the real gate here, since they exercise the from-source build this PR fixes.
User Impact
None on sync behavior — streams, schemas, primary keys, cursors, spec, and state are unchanged. Users get connector images that can actually be built and published again, and the newer CDK underneath.
Can this PR be safely reverted and rolled back?
Link to Devin session: https://app.devin.ai/sessions/b5bc3fe4f6fb4b26b9a95adc0256bae5