Repository navigation
fix: stop calling a failed cloud launch a success - #34844
Conversation
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Fixes #34713 at the cause: the guard now checks how the launch ended. One extra wording branch goes beyond what the Goal needs. Notes
What this change shipsInventory (5 items) — 4 justifiedIntent: a user whose cloud launch failed is told it failed and can retry with changed settings. This is a FIX.
[FIRST-PRINCIPLES-REVIEWED] db7f6e2 Description read[DESCRIPTION-READ] d48e3de40dbf3cc9366ccd7ae1605608540dfd18a205db2ebfaa139f100944cb That sha256 names the PR description this verdict read. Recompute it with |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks. The failed launch now reads the job's outcome, the guard is lifted only when no stack can exist yet, and screenshots of this revision are attached. [DESIGN-REVIEWED] db7f6e2 Description read[DESCRIPTION-READ] 595800bbb7cbea2fc1e5d3c9effb2133e5859fe9bec850e54cd03a26fa0126a5 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
UX Review (Opus 5.5) — ✅ PASSUX-level review of UX-Verdict: PASS No blocker. The blind reader read the new retry line and the re-enabled Launch correctly ("try again … sure"), and every new failure state has a screenshot. Notes
[UX-REVIEWED] db7f6e2 Description read[DESCRIPTION-READ] e21c1672c0c6101b6d43e4f9e07135a016607a9de99224f8db2ba8a3e805cbb3 That sha256 names everything this verdict read: the PR description, plus the 8 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
4ff4dee to
13e95ea
Compare
|
AI review disposition for
|
13e95ea to
438803b
Compare
|
AI review disposition for
|
438803b to
98805b6
Compare
|
UX Review disposition for
|
Code change approver recommendation: Request changesWhy: the decider found 2 problems and estimates an 80% chance that this change introduces a defect, at or above the 80% level where it asks for changes. The main fix works: after a failed launch the line is an inline error instead of the success wording, and Launch is the retry when the launch failed before provisioning. But the new 'crew listed in Your crews' line compares the registry id with the job's EC2 instance id. The registry id is built from the crew's name and the EC2 id is stored as ssm_target, so that line never shows in real use. The test fixture gives both ids the same value, which hides the bug. Frontend lint is also failing on this head, so the new tests never ran in CI. Requesting changes: match on ssm_target and fix the fixture. Problems found:
An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
98805b6 to
42a11d4
Compare
|
AI review disposition for
|
42a11d4 to
7528398
Compare
|
Frontend Tests (3) on |
7528398 to
1c81cb0
Compare
|
AI review disposition for
|
|
UX Review on
|
1c81cb0 to
c964592
Compare
|
Revision
|
After a launch failed, the status line next to the Launch buttons still said the crew was launched and offered "Launch another crew" with the same settings, which would fail the same way. The line now shows an inline error for any failed launch. When the launch stopped before provisioning started, nothing was created, so the same-settings guard is lifted and Launch itself is the retry. From provisioning on the guard stays, since a relaunch could start a second billed crew. The line says the crew is listed in Your crews only when the job registered and its instance is in the live crew list, and otherwise says to check Your crews before launching again. Fixes #34713
c964592 to
db7f6e2
Compare
|
GPT 6.1 finding on
|
|
GPT 6.1 finding on
|
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Tier 1 auto-approve: fix (15 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, security checklist all-NO, no reviewer-lane findings, readiness signal fresh for this head SHA, declared category matches the diff. Category: frontend status fix in the remote crew launch form, so a failed cloud launch no longer reads as a success; adds three translated status lines and tests, with no backend change.
|
Review status: I read the whole diff and have no code concerns. The change adds a failed-launch message, and it lifts main's same-settings guard only when the launch failed before provisioning. That loosens main and adds no refusal. I am holding the approve for one red check. The @JLargent42, please push a commit or edit the PR title or body once. That starts a Sensitive Change Review run for this head, and I will approve once the helper is green. |
Problem / Motivation
Goal: After a cloud launch fails, the line next to the Launch buttons says it failed and offers a retry the user can change, instead of calling it a success.
In Settings → Remote Crew → Set up a new one, after a launch the progress card below shows "Launch failed Kiro Crew Cloud (kc-…)", yet the status line beside the buttons still reads "Launched. Its progress is shown below. To start a second crew with the same settings, choose Launch another crew." Launch stays greyed out, and the only offered action repeats the settings that just failed.
Why it matters
Anyone whose first cloud launch fails (a subnet with no egress, a missing permission) is told it worked and steered to a repeat that fails the same way. The onboarding bug bash hit this in 2 runs. It is the first thing a new user reads after a failure.
Not a goal
What changed (motivation → approach → change)
Symptom: success copy and a same-settings "Launch another crew" after a failed launch. Root cause: the same-settings guard (
alreadyLaunched) compares only the launch body key with the last one sent, and never looks at how that launch ended. So a failed launch counts as a crew that exists.Change in
RemoteCrewPanel.tsx, for the launch these settings went out with (activeJob.id === activeLaunchId) when its job endsfailed:ErrorNotice, never the success wording. It has no hand-off, because the subnet and sign-in choice the next attempt sends live only in this form's state.provisionstep is still pending or skipped, there is noinstance_id, and nothing registered, so nothing exists in the account. Launch itself is then the retry, "Launch another crew" is not shown, and the line reads "Fix the problem shown below, or change the Subnet ID above, then choose Launch to try again." Launch stays enabled with unchanged settings because many causes are fixed outside the form, and the issue's step 3 is "do what it says, then launch again".instance_idis in the liveinstanceslist: "The launch did not finish. Its crew is listed in Your crews, where you can fix or delete it." (launch_failed_crew_listed)launch_failed_check_crews)The three new keys are in every catalog, and
en-XAis regenerated. The retry line names the Launch button by its label in each locale, since main now translates it.Backwards compatibility
Compatible: only the built-in launch form's state after a failed job changes. No API, config or stored data is touched.
Removes nothing: the "Launch another crew" button is still offered after a running or finished launch, and after any failure from provisioning on. It is hidden only after a failure before provisioning, where Launch itself repeats the settings.
Tests
website/src/test/RemoteCrewPanelLaunchForm.test.tsx: new case "after a launch that failed before provisioning, says it failed and offers Launch as the retry". The job is failed atpreflightwithprovisionpending. The test checks that the retry line is an alert, that the success line and "Launch another crew" are not shown, that Launch is enabled, and that a retry with a changed Subnet ID sends that subnet.Same file: new
it.each"after a launch that failed during %s, says it failed but keeps the same-settings guard". The instance list returns the crew only after the launch goes out, as the gateway's would. It has four cases:In each, its own line is an alert and the other line is absent, the success line and the retry line are absent, Launch is off, "Launch another crew" is offered, and a changed Subnet ID re-enables Launch.
All of these fail on main. The listed case fails on the previous head's panel code, since that code never says "listed".
The existing same-settings guard test still passes.
Proof: not provable -- frontend test (prove.py runs Python only)
Local: 58 related frontend test files (
run_scoped_tests.py --surface frontend) pass, plusnpm run typecheck, eslint on the touched files,npm run i18n:check, andtest/test_cloud_sizes_ui_parity.py,test/test_remote_crew_doc.py,test/test_file_line_cap.pyandtest/test_brand_name_gate.pyat-n 2.Manual verification
I rendered the real
RemoteCrewPanelthroughwebsite/capture/remote-crew-update.htmlwith intercepted API answers: one launch whose job comes backfailed, either at the provision step or at the AWS check. I captured it on main and with the fix (below). I did not run a real AWS launch.Screenshots / video
Before (main): success wording and a same-settings "Launch another crew" beside the failed card.
After, failed at the AWS check (nothing created): an inline error beside Launch points at the message and the Subnet ID, and Launch is the retry.
After, failed at sign-in with the crew in the live list: the line says it is listed in Your crews. The card offers its own sign-in recovery.
After, failed during provisioning with no instance: the neutral line says to check Your crews before launching again. "Launch another crew" is the explicit repeat.
After, failed during provisioning, light theme
Related Issues
Fixes #34713
Sibling sweep
Swept: git grep -n "launched_same_settings|alreadyLaunched" -- website/src
Siblings: none -- the same-settings guard and its status line exist only in the built-in form of RemoteCrewPanel.tsx
Pattern harvest
Not generalizable: a one-off guard that keyed on the request sent and never read the job outcome; no second launch guard exists to share a rule with.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributions are accepted under the project's Apache-2.0 license.