Created 02-25-2016 04:56 PM
Hi there,
I am trying to install Spark on 16 node HDP cluster. When I select spark component in the Ambari (Add components page), what happens. Does it install spark library on all the nodes and install Spark server on the master node?
Previously I ran Spark on local mode but want to know how does the installation work under the hood?
Regards.
Created on 02-25-2016 05:04 PM - edited 08-18-2019 05:08 AM
@Smart Solutions you will have to select where you want Spark History Server to run and which machines to install Spark clients on. There is a page further down the road that will allow you to pick which nodes to select what on. For example on Sandbox, if I click either URL below, will take me to sandbox.hortonworks.com, on your cluster, URLs will point to whatever servers you defined.
Created on 02-25-2016 05:04 PM - edited 08-18-2019 05:08 AM
@Smart Solutions you will have to select where you want Spark History Server to run and which machines to install Spark clients on. There is a page further down the road that will allow you to pick which nodes to select what on. For example on Sandbox, if I click either URL below, will take me to sandbox.hortonworks.com, on your cluster, URLs will point to whatever servers you defined.
Created 02-25-2016 06:13 PM
Thanks does it mean we install just Spark clients and nothing related to master processes or server on HDP? While configuring through Ambari?
Created 02-25-2016 06:26 PM
no, you install master services as well. Spark History for example is a service. I don't have access to an installation right now otherwise I'd shown you install steps. Basically components are categorized by clients and master services. You check and uncheck which machines will serve as what. @Smart Solutions
Created 02-25-2016 05:06 PM
Take a look on this
You can pick nodes where you want your clients.
When you install Spark, the following directories will be created:
/usr/hdp/current/spark-client
for submitting Spark jobs/usr/hdp/current/spark-history
for launching Spark master processes, such as the Spark History Server/usr/hdp/current/spark-thriftserver
for the Spark Thrift ServerCreated on 02-26-2016 12:09 AM - edited 08-18-2019 05:08 AM
@Smart Solutions See this
You need master services and clients.
You can see that I have spark client in 3 nodes
when I clicked spark clients and then clicks one of the nodes then you can see all the clients installed in that node
See : Clients
Created on 03-03-2016 05:24 PM - edited 08-18-2019 05:07 AM
When you add Spark through Ambari, you will be asked to choose where to deploy master service (Spark History Service)
And then to choose where to deploy clients services
Finally you will be asked for several properties