0

I'm trying to learn Accumulo. But I have a couple of questions that I couldn't find directly:

  • First, can we add a new server to an existing Accumulo system without any down time? If yes, the new node will have its share (DB data) arranged by master; right? Since it has fail-recovery, I believe that will be automatic.
  • Can we define the number of replications or whole data is shared with some fail recovery system by itself? How can I learn the details of replication and data distribution process?

Thanks a lot :)

Jordan Tigani
  • 26,089
  • 4
  • 60
  • 63
Cihan
  • 13
  • 3

1 Answers1

1

Yes, you can dynamically add/remove worker nodes at any time. They just need to have the same configuration options available to them so that they can join the cluster (shared secret, zookeeper quorum, etc... basically, the same accumulo-site.xml that you are using).

By default, the "master" process will assign tablets to each "tablet server" processes so that each host will be serving roughly the same amount of data.

Not sure I understand your second question, but Accumulo generally uses HDFS for its backing store, which handles replication and data recovery at the "file" level.

MikeD
  • 3,348
  • 1
  • 23
  • 36
  • Thanks a lot MikeD. For my second question, what I wonder is as follows: For example, when we have Mysql Cluster, we define the number of replications we need for the system. I was wondering how Accumulo handles the failures and how many replication it uses. – Cihan Sep 26 '14 at 17:39