* fix(gatekeeper): missing priceId-s should stop the server from booting
* feat(shared): add all new workspace plans
* feat(billing): add new world plans
* feat(ci): use stripe sandbox id-s from test env vars
* chore(ci): remove defunct stripe context
* chore(server-env): fix server env example
* feat(gatekeeper): add missing plans to graphql schema
* test(gatekeeper): update plan upgrade tests to match implementation
* fix(ci): update ci .env source
* fix(ci): we do need some secrets from stripe
* fix(ci): run gql schema introspect with the test env
* fix(frontend): format new plan names
* chore(ci): temp disable gql schema checks
* feat(helm): add unlimited seat prices env vars
* feat(helm): add unlimited seat prices to values
* feat(gatekeeper): do not load priceId-s as a side effect, load with module init
* fix(server/prometheus): ensure consistent Prometheus registry is used
- there was a conflicting dependency between Metrics initialization and Modules initialization; resolved by separating registry initialization from metrics initialization
- pass in the registry to prevent implicit dependency being broken
- when registering a metric, first attempt to remove any of existing metrics with same name to prevent errors
- to prevent sneaky uses of the implicit registry, replace default import with explicit import so it is clearer when prometheusClient.registry is used
* Add tests for registering metrics
- renames corsMiddleware to corsMiddlewareFactory to follow naming convention
- optionally allow cors middleware to be configured
- amend direct calls of `cors` package to now use middleware factory
* feat(multi-region): metrics for knex for all regional databases
* improve typing in knex monitoring
* error logging around migrations
* await async calls for db connections
- add 'region' label
* add missing 'await' statements
* more missing 'await'
* guard against re-adding listeners
* It was possible for update to be called before initialize
- this change collapses both into initialize, and adds checks to ensure initialization is done before being updated for new regions
* separate back into non-exported const and rename
* align with main
* Amend order at which metrics is enabled
* chore(healthchecks): refactor out of modules
* feat(server): multi-region aware liveness and readiness checks
* Add tests for redis & postgres healthchecks
* do not close Redis client after checking it is healthy, we now use a shared client
* userProjectsUpdated.added test
* multi region support
* userStreamAdded
* commit subs tested
* fix for proj workspace assignment
* undo commitHelper main changes
* disable all ffs mode
* createTestWorkspace support when workspaces arent enabled
* project create fix
* workspace projects updated subscription
* WIP new workspace sub
* updated workspaceCreated
* updated workspaceUpdated
* BE for workspace updated sub
* workspace updated sub
* ts err fix
* fix(workspaces): allow workspace delete for paid workspaces
* feat(gendo): multi region gendo with new api integration and limits
* feat(gendo): user credits
* feat: adds limits to gendo panel, and terms and conditions link
* fix: reworks setting back camera view
* feat(gendo): webhook signature verification
* fix(gendo): nullability
* test(blobstorage): fix test dependency injection
---------
Co-authored-by: Dimitrie Stefanescu <didimitrie@gmail.com>
* chore(server): graceful shutdown
- stop() on the apollo server should be called
* chore(server): gracefully drain apollo server
* Allow grace period to be configured
* Terminus manages the readiness and liveness endpoints
* terminus is responsible for stopping the graphql server
* remove logging on shutdown
* Remove redundant parameter
* move healthchecks out of business modules to top-level directory
- terminus can only handle readiness check, not liveness
- app needs to return readiness handler, so that server terminus can use it
* fix tests
* Fix broken merge
* fix broken merge
* incorporate review comments
* fix invalid merge
* fix readinesscheck not being passed as parameter
* fix(file import): server port was broken when localhost was canonical url
- use getServerOrigin() instead of directly calling process.env.CANONICAL_URL
- where server is calling itself, use 127.0.0.1 instead of CANONICAL_URL.
- where server is calling itself (i.e. 127.0.0.1, use the server's listening port and not default for protocol http/80, https/443
- tidy some areas where process.env was called directly; should use envHelper.
* fix tests
- if no data is being sent on a websocket connection then it was closed by nginx after 60s
- it is assumed to be better to keep alive the connection than have the overhead of closing and re-establishing the connection every 60s
Iain Sproat