Reply
New Contributor
Posts: 1
Registered: ‎12-04-2018

NameNode like virtual machine in production

[ Edited ]

 

Hi Everyone, 

 

We are planning to put into production a Cloudera (Impala) cluster with a mix of virtual machines (VMware) and physical hosts (DataNodes), with the following configuration:

 

NameNode Active     (VMware Virtual Machine)

NameNode Standby (VMware Virtual Machine)

PostgreSQL Database (VMware Virtual Machine)

7 x DataNode with Impala Daemon (DELL Physical hosts)

 

We are a kind of new in Hadoop and CLoudera and I am wonder if only putting the NameNode and DB Machine in VMware could dramatically impact the production?  The DataNode will have good resources like 2 Sockets 12 Cores 2.8GHz, 128 GB RAM, 4 disks SATA 10K 1TB, 10Gbps Ethernet connection.

 

Please your comments will be very important for us.

 

 

 

 

Master
Posts: 368
Registered: ‎07-01-2015

Re: NameNode like virtual machine in production

I think the only correct answer for this is to benchmark your setup. I do not have a good experience with mixing VM and PM. But if your IT can guarantee network throughput, latency, and specially IO latency and throughput on all the nodes, then why not?
Announcements