Overview
I had this issue at my customer and wanted to share in case others see it.
Summary:
My customer has a GWLB between its on-prem SAPGUI users (and Amazon Workspaces users) and the AWS SAP App servers. GWLB has a fixed idle timeout of 350 seconds for TCP flows and the customer had set the SAP rdisp/keepalive = 1200, same as their on-prem systems. On-prem routers have a 1 hour idle timeout, so there was no issue on-prem.
When users logged on to SAP running in AWS, they were seeing this:
The SAP dev_disp log showed this:
Wed Apr 27 13:04:53:867 2022
***LOG Q0I=> NiIRead: P=10.53.226.59:49277; L=10.79.14.106:3200: recv (104: Connection reset by peer) [/bas/785_REL/src/base/ni/nixxi.cpp 5500]
*** ERROR => NiIRead: SiRecv failed for hdl 30/sock 16
(SI_ECONN_BROKEN/104; I4; ST; P=10.53.226.59:49277; L=10.79.14.106:3200) [nixxi.cpp 5500]
***LOG Q04=> DpRTmPrepareReq, NiBufReceiveDpTmSend failed () [dpTerminal.c 786]
DpRTmPrepareReq: network error of client T8_U18816 in state DP_LOGGED_IN: NiBufReceive (-6: NIECONN_BROKEN)
DpRTmPrepareReq: client address of T8_U18816 is 10.53.226.59(10.53.226.59)
DpRTmPrepareReq: trigger logoff of T8_U18816
client 000, user 702378829 , terminal DESKTOP-I932URO
DpHdlSoftCancel: cancel request for T8_U18816_M0 received from DISP (reason=DP_SOFTCANCEL_SAP_GUI_DISCONNECT)
NOTE: This is a common question asked by AWS customers. The AWS team is posting it to provide an answer that can benefit everyone.