Repository navigation
fix(daemon): start a lease's TTL when its allocation completes - #3165
Conversation
The registry records a lease, and stamps its expiry, before the lease lifecycle provider allocates the session behind it. Hosted providers can spend 30-80 seconds creating that session, so a lease on the default one-minute inactivity window was expired, or nearly so, by the time its client received it: the next command failed with "Lease is not active", release reported nothing to release, and the paid provider session was left running. The expiry sweep could also reap the lease while the provider was still allocating. Hold a lease work pass for the duration of the provider allocation, the same protection admitted request work gets. The lease cannot expire underneath the allocation, and ending the pass while the requester is still waiting renews the lease for its own TTL from that moment. The response now carries the renewed lease. A requester that hung up still protects nothing, and its allocation is released as before. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
All reported issues were addressed across 5 files
Reply with feedback, questions, or to request a fix.
Re-trigger cubic
… a provider The no-provider allocation test ran on a frozen clock, so it could not tell a preserved TTL from one renewed when allocation completed. On an advancing clock it showed the handler did renew: the lease work pass was held, and its release restarted the TTL, even when no provider allocated anything. Hold the work pass only while a lease lifecycle provider allocates, and run the no-provider test on a clock that advances on every read. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
This PR is ready. At e8f387d the lease TTL now starts when allocation completes, and the handler tests through handleLeaseCommands cover both the provider and no-provider routes. The one reported check passes, but it is a fork head, so the Integration and Coverage jobs may not have run; those need a maintainer to approve the CI run before merge. I did not run the tests or Not blocking, and you can take or leave these: remote-proxy.md:48 says "Either window starts when allocation completes", but on the plain proxy route the TTL is stamped when the record is created, so "when a hosted provider finishes allocating" would be more exact. The The cubic-dev-ai P2 thread on the no-provider test no longer applies, because the test at head holds the pass without a provider; please resolve it: #3165 (comment) |
thymikee
added
the
ready-for-human
…token-attach-c41aff * commit 'd396f3b509d9ed7cddaf170351ea6cf34e02ac16': fix(provider-webdriver): harden BrowserStack app references and endpoints (#3169) fix(android): back off a timed-out snapshot helper session and bound content re-captures (#3160) fix: centralize confirmed daemon retirement (#3126) fix: bind daemon registration writes to the acquired owner (#3125) fix: return confirmed daemon termination outcomes (#3124) refactor(move): share daemon registration and shutdown report modules (#3123) fix: preserve process lock exclusion across publication and reclaim (#3122) fix(cli): refuse a non-URL install-from-source source up front (#3166) fix(daemon): start a lease's TTL when its allocation completes (#3165) fix(android): fail doctor when adb is the Windows binary on a POSIX host (#3157) 0.21.20 0.21.19 # Conflicts: # src/daemon/server/daemon-runtime-metadata-ownership.test.ts
Brings feat/testmu-provider-plugin (915e572) up to date with main, which since landed the split-out callstack#3165 (lease TTL), callstack#3166 (install-from-source URL refusal), callstack#3169 (BrowserStack app references, hub helpers, materialized upload) and callstack#3173 (Limrun attached instances). Main's reviewed versions win for the split-out code: appFileUploadForm with {provider, service}, resolveHubAppReference with parseReference, parseBrowserStackAppReference, isHttpUrl in install, the lease work pass only when a provider allocates, and the upload file check in appFileUploadForm rather than runtime-deployment. Re-applied on top only what the migration needs: provider profile-field refusal (connect, WebDriver and Limrun allocation, refused repeat allocation keeps its lease), optional auth on fetchProviderVerificationJson for TestMu's public catalog, and the provider-webdriver/plugin barrel without asRecord. The TestMu plugin now uses asOptionalRecord from @agent-device/kernel/record, parses lt:// references itself and uploads a public URL before delegating to resolveHubAppReference, and passes {provider, service} to appFileUploadForm. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2 participants
Summary
Split out of #3118. This change affects every provider.
leases.allocateused to start the lease TTL when allocation began. A hosted provider can spend 30–80 s creating its session, so a 60 s lease could expire before the caller got it. The TTL now starts when allocation completes. Two cases are unchanged:A lease allocated without a provider keeps the TTL it was created with. It no longer holds a work pass, which used to renew it on release.
ADR 0007 and
remote-proxy.mdare updated to match.Validation
pnpm check:affected --runpasses.🤖 Generated with Claude Code