Broken pipe error in Cloud Integration

[bestantetholits]

Since the start of June we are seeing failed messages for our integration with error: javax.servlet.ServletException: Error during write of the response: , cause: java.io.IOException: error 32 - Broken pipe.

Though the messages failed in Cloud Integration we can confirmed that the file was delivered to the webservice endpoint successfully. This issue is recurring though there are cases of the same integration flow executions that were successful in Cloud Integration. So the issue is intermittent. Last update of the integration package happens on May 13 so we have no idea what is causing this error. Just to give more context on the integration flow, the integration flow is called via Integration Center as REST API URL and is triggered via an event in ISC (Intelligent Service Center) in SF. We tried replicating it in our internal system but we cannot replicate the same issue reported by our customers.

[bestantetholits]

maik.offerle thank you! We have a ticket to SAP. See below:

They are claiming that we need to do something at the CPI end. Can you let us know where in SF you set or configure how long the http session is open to CPI? Thanks!

[bestantetholits]

Hi maik.offerle,

Sorry for the delay in my response. Can you provide further info on how to achieve your suggestion: " higher timeout for synchronous HTTP sessions on the sender system side"?

Thanks!

[maik_bosch]

this is nothing which can be done on CPI side. It is a setting on sender system side. They have to check how long they keep the http session open to CPI. It seems like on their side they only leave it open for 3 minutes. They should change it to 5 minutes or more. See some discussion about advantages and disadvantages here: https://stackoverflow.com/questions/33712208/maximum-reasonable-timeout-for-a-synchronous-http-reque...

[bestantetholits]

Hi maik.offerle thank you for your response. Sorry for the delay in my feedback. See screenshot of what we have in Integration Center monitor which corresponds to the message in Cloud Integration which completed successfully.

Here's the screenshot of the one that triggers a failed message in Cloud Integration. Based on the time stamp it is taking 3 minutes+ before the http error is raised.

This is what you can see in Cloud Integration where the soap call executed successfully but the iflow failed because of that error read time out:

Any suggestion on how we can resolve this? This is working fine since 2019 and we just had this issue late 3rd week of May this year. Thanks!

[maik_bosch]

Hi angelitobestante ,

as you can see from the error message the issue is happening during write of the response to the sender system

Means the connection was already closed by the sender system when CPI tries to give the sender back the response

That's why you also see the file transferred successfully.

So it could be solved if you put a higher timeout for synchronous HTTP sessions on the sender system side...

Or you handle the exception (which is not causing real issues?! - or does the sender system rely on a response?) in an exception subprocess.

BR

Maik