I had a three-node Brisk cluster.
node 1 : 10.64.1.1
node 2 : 10.64.1.2
node 3 : 10.64.1.3
I choose the seed node (node 2) as jobtracker.
I can see the whole cluster using nodetool.
From http://10.64.1.2:50030/jobtracker.jsp, I can see this three nodes.
Today, on node 1, I used "brisktool movejt 10.64.1.1". That said, I change jobtracker to node 1. Node 1 becomes a single hadoop cluster. After I finished the mapper/reducer job on node 1 (I submitted a mapper/reduce job in Hive), I used "brisktool movejt 10.64.1.2".
That said, I want to change back the jobtracker. However, it does not work. It does not return to three-node hadoop cluster, although it is still a three-node cassandra cluster.
From http://10.64.1.2:50030/jobtracker.jsp, I can only see two nodes.
On node 1, /var/log/cassandra/system.log shows the following errors:
INFO [TASK-TRACKER-INIT] 2011-07-26 20:19:37,000 Client.java (line 629) Retrying connect to server: 10.64.1.2:8012. Already tried 9 time(s). INFO [TASK-TRACKER-INIT] 2011-07-26 20:19:37,000 RPC.java (line 321) Server at 10.64.1.2:8012 not available yet, Zzzzz...