Repository navigation
Fix billing adapters and add a reproducible GKE/Stripe example - #3
Merged
Merged
Conversation
Automatic subscriptions refused such a plan with a retryable error, so the pipeline retried that tenant forever and held up Stripe delivery for every tenant. The error is now permanent: the tenant is logged as rejected, its customer and usage keep flowing, and everyone else is unaffected.
A rate card change created a contract on the new rate card and left the old one open, so usage was rated twice. vBilling now keeps one open contract per tenant: the new contract starts at UTC midnight and every other open vBilling contract of the tenant ends there, or is archived if it would not have started by then. Contracts vBilling did not create are never touched. Uniqueness keys carry the start date (vbilling:<rate card>@<date>:<tenant hash>), so a tenant can return to an earlier rate card; v0.2.0 keys are recognized and those contracts adopted. A tenant that loses its state adopts its open contract instead of creating another. With the Stripe link on, customers that vBilling did not create (found by alias or pinned by ID) now get the Stripe billing provider configuration when they have none. The new requests are validated against Metronome's published OpenAPI schemas, added to testdata.
Billing platforms report totals per tenant and metric, but a sku, region, project or cluster filter summed only the matching ledger events, so a tenant with two SKUs always mismatched. Filters now pick the tenant and metric pairs to compare, and each row covers all of that tenant's usage of the metric.
PUT /api/v1/billing-states/{tenant} returned 404 unless a Stripe or
Metronome webhook secret was set, so Lago and custom billing platforms
could not drive enforcement. The enforcer now always exists; webhooks
still need their own secrets.
- A webhook receiver that rejects the catalog event with 400, 413 or 422 no longer pauses usage delivery: the catalog is informational. - Offboarding records each destination's result: a permanent rejection is final, and destinations that succeeded are not called again while another one retries. - Lago treats 404 on terminating a subscription as already done, and its DELETE errors are classified like its other requests.
Stripe plans over 20 prices, Metronome contract handover and Stripe links, filtered reconciliation, and billing states without webhook secrets. Also corrects the gpu_profile example in a code comment.
saiyam1814
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This change fixes billing adapter failures that could stall usage delivery or produce incorrect reconciliation results, and adds a reproducible GKE GPU example with Stripe test billing.
This PR builds on the adapter guides merged into
mainin PR #2.Billing behavior
Reproducible GPU example
examples/gke-stripedocuments shared L4 GPU allocation, an optional private A100 worker, the native Go collector, customer/meter mapping, Stripe Products/Prices/subscriptions, usage verification, recovery and scoped cleanup. Helpers keep credentials and runtime state outside Git, restrict Stripe operations to test mode, and validate resource ownership before deletion. CI runs the offline helper tests and a focused credential-pattern check.CI reliability
Dispatcher and API tests now wait for their background dispatcher to exit after cancellation, including failure paths. This prevents final cursor writes from racing with spool and temporary-directory cleanup. Production Go behavior is unchanged by this test-lifecycle fix.
Validation
go vet ./...go test -race ./...go test -race ./internal/pipeline ./internal/api -count=100Adapter tests cover the billing fixes, including Metronome request schemas derived from its published OpenAPI specification. The Metronome changes have not been tested against a live Metronome account. Enforcement for tenant clusters with their own nodes remains a documented limitation.