Cannot create a HA cluster (active and standby name node) using Ambari blueprint. - Ambari 2.2.1.1

[#1]

Wanted to set up HA (active name node and standby name node)cluster. We did not want to have a secondary name node to be present. Just two name nodes and one of them will be active and the other as standby. Used the Ambari blueprint exactly as outlined in the link: https://cwiki.apache.org/confluence/display/AMBARI/Blueprint+Support+for+HA+Clusters

Getting an error:

{\n "status" : 400,\n "message" : "Cluster Topology validation failed. Invalid service component count: [SECONDARY_NAMENODE(actual=0, required=1)]. To disable topology validation and create the blueprint, add the following to the end of the url: \'?validate_topology=false\'"\n}

Tried to disable topology validation with validate_topology=false, Blueprint registered but while creating cluster creation failed with the error given below

java.util.concurrent.ExecutionException: java.lang.Exception: java.lang.IllegalArgumentException: Unable to update configuration property 'dfs.namenode.https-address' with topology information. Component 'NAMENODE' is mapped to an invalid number of hosts '2'.

Any pointers to sort this out will be very helpful.

Thanks,

Mohan

[#2]

Hi Mohan,

Have you started from the example blueprint from https://cwiki.apache.org/confluence/download/attachments/55151584/hdfs_ha_blueprint.json?version=4&m... ?

You'll need to have a ZKFC component where you have the NAMENODE component:

        { "name": "NAMENODE" },
        { "name": "ZKFC" },

And 3 of JOURNALNODE:

        { "name": "JOURNALNODE" },

If that's the case, can you attach your blueprint and cluster creation template please?

Alex

[#3]

hdfs-ha-blueprint.txtclustertemplate.txt@Alexandru Anghel: Thanks for a quick revert. Saw the link that you posted. It is something similar to the ones that we had tried. Yes, the ZKFC component is installed in the same host where the NAMENODE component is installed. Attaching two files: blueprint and the template

[#4 · Accepted solution]

Hi Mohan,

Looks like you're missing the following from hdfs-site properties:

"dfs.nameservices" : "mycluster"

Which is quite important as it defines the HA HDFS logical name.

Hope it works once you add it.

Alex

[#5]

Hello Alexandru, Great. Think that it should be the reason - 100% :-). Will verify the same and get back to mark this as an answer. Your quick pointer to the issue has been really really helpful. Thank you so much.

brgds, Mohan

[#6]

Hello Alexandru, It worked. Thank you.Now, I am checking for similar properties for Resource Manager, Hive, Oozie to make them highly available from the blueprint itself rather than creating the cluster using Ambari and then manually making it HA for those services. Thanks, Mohan