Hi Team,
We have multiple SAP systems running on VMware ESXI most of them are Linux and only one windows system where SAP PI is running (Windows 2012 R2).
There is a strange issue occurring in the system like when multiple requests sent to this windows VM without any issues all packets are received. However once the received packets completes its process in SAP PI system then try to send (transmit) to the same target system (Linux) a network congestion happens and lot of TCP Retransmission occurs and many of the connections goes time out.
With wireshark trace and Microsoft message analyzer we can observe lot TCP Retransmissions happened in the system and eventually it failed processing.
We have already tried to set the recommendation from the below link but still it dosent help. Also during the network congestion only the new TCP transmission gets time out error and established sessions works fine.
https://support.microsoft.com/en-us/help/170359/how-to-modify-the-tcp-ip-maximum-retransmission-time...
On further analysis we found that the issue is only with VMXNET3 adapter, because any changes at the OSlevel there is no effect in the performance. The network congestion happens only if there is bulk data transmission. However when we check the amount of data sent and received, it is only 1/3 of the actual that we need to perform in the system. As well as we have a physical host which is a copy of same windows system which works perfectly even during full load.
Please let us know what are all the parameters that needs to be checked and tweaked under this kind of situation on VMXNET3 side.
Note:- we did a thorough investigation with network team and found there is no issue on the network as both the VM's communicating are in single VLAN and there is no firewall in between. We have even tried to do Vmotion for both VM's and made them reside in single ESXI host. But still same network timeout occurs on the windows VM.
Regards,
Giridharan
