Question Re: 504: GATEWAY TIMEOUT error in Cloud Foundry Environment in Technology Q&A

504: GATEWAY TIMEOUT error in Cloud Foundry Environment

elamparithi666 — Tue, 09 Jun 2026 09:19:18 GMT

When calling the API in the SAP BTP Cloud Foundry (CF) environment, it connects to the backend OData service (on-premises) to fetch data from the SAP system. The target endpoint "default" parameter io.timeout.millis has been configured to 60 seconds as per the SAP documentation. However, the issue is that a "504 - Gateway Timeout" error is returned by the API before 60 seconds.
Why is the io.timeout.millis configuration not being considered or meaningful in the CF environment? Could this be related to a limitation or issue in the SAP BTP Cloud Foundry environment?

https://help.sap.com/docs/sap-api-management/sap-api-management/target-endpoint-properties

Re: 504: GATEWAY TIMEOUT error in Cloud Foundry Environment

oliver_maynard — Mon, 06 Jul 2026 11:07:15 GMT

Dear elamparithi666, your question appears to concern Cloud Foundry runtime. I recommend changing the SAP-Managed tag to "SAP BTP, Cloud Foundry runtime and environment". Kind regards, Oliver

Re: 504: GATEWAY TIMEOUT error in Cloud Foundry Environment

furkan5 — Wed, 14 Oct 2026 14:44:41 GMT

Hi, According to the API Management documentation you linked, the default timeout is 54 seconds for APIs created based on an on-premise backend. An explicitly configured `io.timeout.millis` value is applied only when it is lower than the applicable default. Therefore, setting it to `60000` does not extend the timeout to 60 seconds in this scenario. [SAP documentation](https://help.sap.com/docs/sap-api-management/sap-api-management/target-endpoint-properties) However, we still need to confirm whether this explains your 504 error. **After exactly how many seconds does the error occur, and what error code appears in the API Management trace?** If it occurs around 54 seconds, that is consistent with the documented limit. If it occurs significantly earlier, another timeout in the request chain should be investigated. If the backend OData operation takes longer than the supported timeout, I would consider optimizing the query and reducing the amount of returned data. An asynchronous flow may also be appropriate for long-running operations.