So-called Edge nodes provide a common user access point for the MapR webserver and other client tools. Edge nodes may or may not be part of the cluster, as long as the edge node can reach cluster nodes. Nodes on the same network can run client services, MySQL for Metrics, and so on.
However, I'm still not clear.
a) Why need 'provide a common user access point for the MapR webserver' ?
b) If I want to read/write to MapR-FS. It is just NFS mount point it .. or it is necessary to go thru Edge Nodes?
Thanks in advance.
Answer by Nabeel Moidu · May 23, 2013 at 09:46 AM
a) So that all the functionality that directly doesn't involve core cluster services like Fileserver/Tasktracker/Jobtracker/HBaseRegionserver/HBasemaster etc can be put into a single node and can be a single point of access. Otherwise you would have identify where the webserver is running, where the mysql for metrics is running , where other services are running etc.
On the actual cluster nodes, you can devote the full resources on the node for actual filesystem i/o, mapreduce processing and hbase etc. This isn't mandatory. It's just a suggestion for better architectural planning.
b) If you want to read write via NFS, then host a MapR NFS server either on a cluster node or on the edge node. Then point the nfs client to mount it. Having an mapr NFS role on any one node is mandatory for NFS mounting the filesystem. It's not a must to have it on the edge nodes only.
Answer by lc2a · May 23, 2013 at 09:58 AM
If not planning to use NFS, then It is the Edge node will just install with MapR Client http://www.mapr.com/doc/display/MapR2/Setting+Up+the+Client#SettingUptheClient-client
It is need to have cluster of Edge nodes? for fail-over? What will be the configuration so that each Edge nodes awares & know it belong to the Edge cluster?
Answer by praveenshettigar · May 30, 2013 at 06:04 AM
Hi Nabeel - b) If you want to read write via NFS, then host a MapR NFS server either on a cluster node or on the edge node. Then point the nfs client to mount it. Having an mapr NFS role on any one node is mandatory for NFS mounting the filesystem. It's not a must to have it on the edge nodes only.
From what you are saying the ability to read and write can be achieved by using an NFS client on the edge node where edge node is outside the cluster or make edge node part of the cluster and have MapR NFS Services run on it.
Is there a performance gain if I use MapR NFS service on the edge node versus an NFS client on the edge node. I am trying to understand if there is an advantage of one aproach over another.
Thank you! Praveen