AnsweredAssumed Answered

Correct way to rollout a 5 node cluster with ZK high availability

Question asked by rhinomike on Feb 23, 2015
Latest reply on Feb 25, 2015 by leonclayton
Hi there,

I am playing with the M3 version of Mapr and I seem to have hit a blocker.

I managed to get 5 nodes running with success, on a setup like:

* node 1 runs ZK, CLDB, NodeManager, NFS, Webserver, HBase Master, Fileserver, RM, HIstoryServer and NodeManager
* node 2 runs NodeManager, FileServer, HBase Region Server
* node 3 runs NodeManager, FileServer, HBase Region Server
* Node 4 runs NodeManager, FileServer, HBase Region Server
* Node 5 runs NodeManager, FileServer, HBase Region Server

Till here all good, system work as planned without surprises, however, when I try to promote Node 4 and 5 to backup ZK and CLDB it is hell on earth and the whole thing collapses.

I suspect that it is somehow connected to the use of the new configure.sh from:

./configure.sh-Z node1 ...

to

./configure.sh-Z node1,node4,node5 ...

My understanding is that indeed I should connect to every single node in the cluster **including the ZK nodes** and invoke configure.sh with the new ZK nodes under the -Z option. **Is this correct?**

If not, could someone point me in the right direction?

I thank you in advance

Outcomes