Skip to Content

PI Unable to create a socket

Hello, dears

I know, that this question has already been asked many times. None of the existing answers could help me. I have been searching for the solution for about 6 months.

We have several simple PI scenarious. The third party calls synchronous SOAP web-service, which is replaced on ERP side. So, we have Caller side -> PI -> ERP.

Everything worked fine for a long time. We suppose, that the problem appeared, when the load on system has increased (many services calls).

Now, we have the fllowing situation. All such services in one of our several ERP systems (we have several ERPs, but such problem is only in one of them) periodically failed with error "Communication over HTTP. Unable to create a socket.".

This is not a network problem, because sometimes the service is called fine, and sometimes it fails. And it failes many times a day. Maybe, about 50% of calls are failed.

The basis team said, that they configure the ERP system to get a huge number of connections, but it doesn't help.

I will be very appreceated for any help.

Thanks a lot.

Add comment
10|10000 characters needed characters exceeded

  • Get RSS Feed

3 Answers

  • Jul 03, 2017 at 12:57 PM

    Hi Ekaterina!

    Maybe this blog could help:

    https://blogs.sap.com/2011/01/26/tuning-the-pi-messaging-system-queues/

    Regards, Evgeniy.

    Add comment
    10|10000 characters needed characters exceeded

  • Jul 03, 2017 at 02:23 PM

    Hello,

    Do you have any other SAP system like SCM linked with your ECC? Have you verified what are the processes that were running during the same interval? This could be due to heavy load processing in ECC and not able to accept new messages.

    Another thing, can you verify if the issue is related to 1 specific cluster node may be. Try to see the failed message and try to check if it was related to 1 specific cluster may.

    Cheers!

    Hari

    Add comment
    10|10000 characters needed characters exceeded

    • Hello, Hari

      Yes, it could be due to heavy load processing in ECC. I have written about it in my answer ("We suppose, that the problem appeared, when the load on system has increased (many services calls)."). But can we do something with it?

      I have checked, the issue is appeared on all cluster nodes (two in our case).

  • Jul 10, 2017 at 11:41 AM

    Hello! Does anybody know the answer? Or maybe somebody has some useful information? I will be very appreceated for any help.

    Add comment
    10|10000 characters needed characters exceeded