We have a gateway which is acting as a LORAWAN concentrator with few LORAWAN devices.
When we see the data we could able to see the gateway ID from which data is transmitting.
Now we powered up another gateway within the vicinity of device, all the data to TTN is getting routed via new gateway. But that gateway is not configured to the user profile that respective user.
Any suggestions how to disable this? we want the device to communicate to TTN via initial gateway only.
Devices just transmit it’s packet via RF, they are oblivious if there are any gateways or not receiving there RF…
Gateways receive packets via RF, the are oblivious bout who the node is.
As long as the gateway is registered on TTN and have a active connection to the NS it will forward the packets. (as long as it pass CRC test)
The NS looks and make decisions about the packets, like to what application it belongs to etc.
The short of the long story is that all TTN gateway will receive the packets they can ‘hear’ and pass it onto the NS. You can’t set what node goes where.
Broadly as Johan described as LoRaWAN is essentially a broadcast technology and the LNS will be reporting the signal with the best quality - typically the one received first(*) and the strongest RRSI etc.
(*) A GW closer to the device that sits on a slower backhaul may end up not being the primary reporting GW if one further away is on a faster ping time to the LNS as its messages will reach the GW 1st, even in the RSSI for that GW is a few db lower due to the increased distance.
What is key, however, is that if you look at the meta data for the message it will include ALL gw’s that received and forwarded the message to the LNS, provided it wasnt delayed too long through backhaul routing such that its late arrive is out side the preset LNS reception window and doesn’t then end up looking like a replay attack! 
The good news therefore is that you can capture the full metadata for each message - e.g. through your own webhook(s), in your own back end application/database, filtering and then presenting the data to your user based on just the metadata delevered again your target GW… sounds like you have a bit of coding to do in you application server 
One thing else you need to consider now you have a fail over now for the nodes, as you know the nodes RF can be received by the ‘other’ gateway.
If these gateways are on separate network even better. Then you don’t not only have redundant gateway but paths to the NS as well. 
is there any way like whitelisting, where I want X no of devices to communicate via gateway 1 and other set of Y devices communicating to Gateway 2.
This problem becomes more prominent where I want to perform the range test of the IOT device w.r.t gateway and I dont have control to power down other gateways.
Thanks
Short answer no. White listing (and blacklisting) would break TNN Manifesto and FUP. Even then would not work as you are not in control of the ‘foreign’ GW you have no access to!.
As above not an issue - use the metatdata as above!
I have deployed a great many GW’s both under my own user ID and for collaborators and clients across many territories and lots of TTN Communities and I find myself mapping literally every day - trackers/mappers in all vehicles, carried when out for walks/exercise etc. So if i need to look at a specific device/GW relationship that is easily done if I look at the GW’s I have access to (from back end point of view), or via reported out metadata through various dashboards and tools like e.g. TTNMapper etc.