how sqoop talks to external world.

[#1]

We have a cluster of 28 nodes and working in a firewall Environment and scoop client is installed in all 28 nodes. We need to scoop data from oracle(outside organisation) to hive. For Sqooping do we need to open all the ports 28 nodes IP to external oracle IP. or is there any alternative way by which we can give a single IP and establish the connection.

[#2 · Accepted solution]

@Gaurav, If you are trying to use SQOOP CLI, then from the server,where you are triggering,your firewall should be exposed with oracle.

But if you are trying to use through oozie, since you have installed sqoop client in all the 28 servers, so oozie can pick any, so firewall should be opened from all the nodes.

[#3]

Thanks Rajendra, as of now we have installed sqoop client in 28 nodes, so can we uninstall the sqoop client and keep the client in 2 or 3 nodes ??

[#4]

As it is installed on all the nodes, firewall should be opened from all 28 nodes, else restrict the sqoop client to limited number and open the firewall from that.

[#5]

No Gaurav, let it be installed on all 28 nodes as it is now. Because oozie will not be sure to pick which node, it can be anyone out of 28. So now firewall should be opened from all nodes. Please clarify Rajendra if my understanding is wrong.

[#6]

@Vikram Kumar Katilu

You are right.