Changelog

Behaviour changes

• Introduce a timeout while processing Event Triggers. The timeout will be determined by the lower value, 30 minutes or the response timeout associated with the Event Trigger.

Bug fixes and improvements

Server

• Add support for pg_get_event_logs, pg_get_event_invocation_logs, pg_get_event_by_id, mssql_get_event_logs, mssql_get_event_invocation_logs and mssql_get_event_by_id APIs. See the docs for more details.
• Throw source-cleanup-failed warnings when an inconsistent source is removed using the replace_metadata or clear_metadata metadata APIs only if there are event triggers associated with the removed source.
• Mark Event Triggers as inconsistent if GraphQL Engine is unable to create the Event Trigger infrastructure (SQL triggers and tables for logging) on the source.
• Add a new Prometheus metric hasura_websocket_message_queue_time and extend the hasura_graphql_requests_total to also include subscriptions requests. (Cloud / Enterprise edition only)
• Add a new Prometheus metric hasura_active_subscription_pollers_in_error_state for better observability in the subscriptions subsystem. (Cloud / Enterprise edition only)
• In the OpenTelemetry exporter, set the service.name resource attribute to "hasura" by default. (Enterprise edition only)
• Improve tracing spans and OpenTelemetry exporter. The top level trace is annotated with request_id and parameterized_query_hash. Bulk queries should output a separate parameterized_query_hash for each query span. (Enterprise edition only)
• Fixed checking of remote schema to remote schema joins in the caching restriction check. Hasura doesn't allow queries with remote schemas to be cached when forward_client_headers is set. But remote schema to remote schema joins were not checked for this restriction. This is now fixed. (Cloud / Enterprise edition only)

Console

• Updated create REST endpoint form to fix the base URL from a fixed label of localhost:8080 to the actual server URL.
• Add warning badge for non-cached queries with warning response header in API Explorer.
• The "Track All" action for tables in Data Connector-backed databases does not fail for the entire batch if there are conflicts with existing schema for a one/more tables. The ones that can be tracked will be tracked successfully.
• Clicking the name of a tracked table in the new table tracking UI will now link to the manage table page.
• Add Console support for warning notifications while importing and resetting Metadata, and setting API time limits.
• Fix a UI bug in inspect modal in the operation tab under monitoring. The modal was broken if any error was associated with the inspected operation. (Cloud / Enterprise edition only)
• Enable SSO token authorization from the Console for Enterprise Edition. (Enterprise Edition only)