There are a couple sources for information I'll highlight here in hopes that they may be useful. I don't have a direct answer but perhaps with more information and a bump to the top of the thread list, more people may respond.
The Administration guide for OneFS also had information on configuring the Ambari agent it's on page 885, but I've pulled out the text below. There does not appear to be any connection with SmartConnect, just the Access Zone. So the connections should follow the appropriate assignment for that Access Zone.
1.In regard to automating, yes the Isilon integration into ambari is different than other hosts, it doesn't use ssh keys to allow the ambari server to connect/install the amabri agent on the isilon. Rather the Isilon reverse registers itself into the ambari server once you have configured it on the Isilon side(it's sort of reverse, the Isilon is trying to connect to ambari server itself). The process i've seen is to use the ssh keys method to connect to and bring hosts into the amabri setup BUT then the Isilon sort of just shows up as it is constantly attempting the register itself against ambari which the setup picks up on.
Once the Isilon is in the ambari setup you just select the roles for it; namenode, secondary-namenode and datanode and select all other roles for other hosts; YARN/HIVE/zookeeper etc etc. Unfortunately i have no experience on how to automate this or if this will create additional work. But it maybe that if the Isilon host is present in an automated install it takes care of registering itself with ambari and the install can continue.
2. From page 439 of the OneFS 7.2 Admin Guide: "The NameNode must be a resolvable SmartConnect zone name or an IP address from the address pool associated with the access zone." So yes technically you can set this up without using a SmartConnect Zone name but by doing so you very clearly limits the scalability and resilience of your isilon services and namenode connections, using a SCzone name allows you to have multiple interface from multiple nodes participate in the pool (the SSIP acting as the point of entry into the pool) . It is strongly suggested to use SCzone names for client connections, so connections are not tied to only a specific IP on a single host. If using a dedicated hdfs access zone a corresponding pool with a SmartConnect zone should be used.
cadiletta
1 Rookie
•
106 Posts
0
January 22nd, 2015 07:00
There are a couple sources for information I'll highlight here in hopes that they may be useful. I don't have a direct answer but perhaps with more information and a bump to the top of the thread list, more people may respond.
Ambari Agent
The Administration guide for OneFS also had information on configuring the Ambari agent it's on page 885, but I've pulled out the text below. There does not appear to be any connection with SmartConnect, just the Access Zone. So the connections should follow the appropriate assignment for that Access Zone.
https://support.emc.com/docu56048_OneFS-7.2-CLI-Administration-Guide.pdf?language=en_US
Configure Ambari agent settings
You can configure Ambari agent support in each access zone that contains HDFS data.
Ambari agent settings can only be configured through the command-line interface.
Procedure
1. Open a secure shell (SSH) connection to any node in the cluster and log in.
2. Optional: To identify the name of the access zone that you want to configure for
Ambari agent settings, run the following command:
isi zone zones list
3. Run the isi zone zones modify command.
The following example specifies company.ambari.server.com as the external Ambari
server that receives communication from the Ambari agent running in the zone3
access zone:
isi zone zones modify zone3 \
--hdfs-ambari-server=company.ambari.server.com
The following example designates the IP address 192.168.205.5 as the point of
contact in the zone3 access zone for Hadoop services managed through the Ambari
interface:
isi zone zones modify zone3 --hdfs-ambari-namenode=192.168.205.5
russ_stevenson_
25 Posts
0
January 27th, 2015 09:00
With regard to the two questions:
1.In regard to automating, yes the Isilon integration into ambari is different than other hosts, it doesn't use ssh keys to allow the ambari server to connect/install the amabri agent on the isilon. Rather the Isilon reverse registers itself into the ambari server once you have configured it on the Isilon side(it's sort of reverse, the Isilon is trying to connect to ambari server itself). The process i've seen is to use the ssh keys method to connect to and bring hosts into the amabri setup BUT then the Isilon sort of just shows up as it is constantly attempting the register itself against ambari which the setup picks up on.
Once the Isilon is in the ambari setup you just select the roles for it; namenode, secondary-namenode and datanode and select all other roles for other hosts; YARN/HIVE/zookeeper etc etc. Unfortunately i have no experience on how to automate this or if this will create additional work. But it maybe that if the Isilon host is present in an automated install it takes care of registering itself with ambari and the install can continue.
2. From page 439 of the OneFS 7.2 Admin Guide: "The NameNode must be a resolvable SmartConnect zone name or an IP address from the address pool associated with the access zone." So yes technically you can set this up without using a SmartConnect Zone name but by doing so you very clearly limits the scalability and resilience of your isilon services and namenode connections, using a SCzone name allows you to have multiple interface from multiple nodes participate in the pool (the SSIP acting as the point of entry into the pool) . It is strongly suggested to use SCzone names for client connections, so connections are not tied to only a specific IP on a single host. If using a dedicated hdfs access zone a corresponding pool with a SmartConnect zone should be used.