Tyk Gateway v5.2.1

Fixed

• Fixed a memory leak issue in Gateway 5.2.0 if opentelemetry is enabled. It was caused by multiple otelhttp handler being created. We have updated to use single instance of otelhttp handler in 5.2.1 to improve performance under high traffic load.
• Fixed a memory leak that occurred when setting the strict routes option to change the routing to avoid nearest-neighbour requests on overlapping routes (TYK_GW_HTTPSERVEROPTIONS_ENABLESTRICTROUTES)
• Updated panic message from Spanish to English
• Fixed a bug where missing http.response.body.size and http.request.body.size attributes to both Tyk HTTP span and upstream HTTP span. It is to allow users to have visibility on request and response size from the traces.
• Fixed a bug where the Gateway did not correctly close idle upstream connections (sockets) when configured to generate a new connection after a configurable period of time (using the max_conn_time configuration option). This could lead to the Gateway eventually running out of sockets under heavy load, impacting performance.
• Fixed a bug where HTTP APIs, when Persist GraphQL middleware was configured in conjunction with Response body transform, the body transformation of the response was not applied.
• Fixed a bug during Gateway reload where the Gateway would continue to load new API definitions even if policies failed to load; this would leave the customer at risk of an APIs being invoked without the correct policies having been loaded. Now Tyk will make a configurable number of attempts to load resources (APIs and policies) before continuing. If a resource fails to load, an error will be reported in the log and the Gateway will cancel the reload reverting to its previously working resource configuration.
  We have introduced two new variables to configure this behaviour:
  resource_sync.retry_attempts defines the number of retries that the Gateway should perform during a resource sync (APIs or policies), defaults to zero which means no retries are attempted
  resource_sync.interval configures the fixed interval between each retry attempt (in seconds)
• Remove the extra chunked transfer encoding that was added to rawResponse analytics
• Fixed a bug where, if you created a Key which provided access to an inactive or draft API, you would be unable to subsequently modify that Key (via the Dashboard or directly via the Tyk Gateway API)
• Fixed a potential performance issue related to high rates of Gateway reloads (when the Gateway is updated due to a change in APIs and/or policies). The gateway uses a timer that ensures there at least one second between reloads, however in some scenarios this could lead to poor performance (for example overloading Redis). We have introduced a new configuration option reload_interval that can be used to adjust the duration between reloads and hence optimise the performance of your Tyk deployment.
• Fixed a bug where headers not being passed upstream for GQL/UDG subscriptions

Tyk Dashboard v5.2.1

Fixed

• Fixed a bug where "Add GraphQL Operation" tickbox could not be disabled in GQL data source configuration screen, even if no operation was added. Now it's possible to change its state depending on the presence of GQL operation and variables.
• Fixed a bug in the Dashboard UI where pagination was breaking for lists using GraphQL requests
• Fixed the Tyk Dashboard API so that Tyk Sync can fully support Tyk OAS APIs; this will be enabled from Tyk Sync version 1.4.1.