Hbase stopped replication

Could you please respond?

Are you talking of HDFS replication or the HBase's own event replication?

Also, you may want to ask CDH specific questions at
cdh-user@cloudera.org lists instead of the SCM lists which is for CM
questions and has such an audience.

Earlier, in Hbase, if I created a table, it used to create it in 2
nodes.(Ours is 2 node cluster).
Now it does not do that.

Thanks for you help!

What do you mean by that exactly? Are you seeing it not replicate data
across datanodes (how are you noticing that, fsck?) or something else
in terms of 1 region vs. 2 regions?

It does not replicate across datanotes.
I have 2 nodes in my cluster.
Namenode and Regionserver1 is in node1 and regionserver2 is in node2.

Using the Cloudera Manager, I click on
hbase service -> regionserver1 -> Hbase region server web UI
(http://mynode1:60030/rs-status)

I do not see my table.

Repeat the same steps for node 2
hbase service -> regionserver1 -> Hbase region server web UI
(http://mynode2:60030/rs-status)

I see my table.

Also, using the CM, when I click on hosts, I see node1 and node2. The disk
space used by node1 is 28GB and node 2 is 12GB.
The sizes of both used to be almost same.

I noticed that balancer service is not running. Replication problem is
because of that.

Thanks

Hi Mike,

HBase's regions and blocks it uses on HDFS are two wholly different
concepts - do not confuse between the two.

Tables will be shown on the Master web UI of HBase alone. On
RegionServer Web UIs, you see only regions that belong to a certain
table. If you have just one region in your table then only one
regionserver can serve it and it would alone show it up when you
check. If you want this to be "2", then you have to split the region
appropriately such that there are more than 1 region in your table and
each is served by HBase's balancer from an individual RS. The regions
have nothing to do with the underlying HDFS block replication for
data, which is still handled fine.