Hi,
Regarding your questions:
Q- what should the source IP will be hitting the actual server ? (STEP 2)
A/ Assuming that you are not doing NAT in your LB, the source IP on the server is going to be the original client IP: 10.1.1.1.
Q- I have also seen we get the actual source initiating the connection hits the server how would that works ?
A/ In that diagram, there are 2 ways for that to happen, 1) the server is using the LB as the default gateway, so all the traffic that is not from the local subnet, is going to be sent to the LB as the next hop, and that will complete the flow on the LB. 2) there is another L3 device on the server subnet either using routing or PBR in order to send the reply back to the LB.
Q-If souce nat is configured and client reside in same vlan as real servers how would session table look like ?
A/
1 | Client -> LB | 192.168.1.50 | 192.168.1.200 |
2 | LB -> Server | 192.168.1.201 | 192.168.1.11 |
3 | Server -> LB | 192.168.1.11 | 192.168.1.201 |
4 | LB -> Client | 192.168.1.200 | 192.168.1.50 |
Where 192.168.1.201 is my NAT IP.
Regarding the X-Forwarder-For, usually this is used when you are using NAT, so the client IP is included on the HTTP headers, so the servers can keep track of the client IP's, but it depends on the LB and if the feature is supported, also you need to take into account the protocol you are load balancing, it works for HTTP.
Hope it helps.