Sorry, here's what we are using:
* All 3 nodes are m1.large ebs instances.
* 1 Master node running the following services:
o hadoop-0.20-namenode
o hadoop-0.20-secondarynamenode
o hadoop-0.20-jobtracker
o hbase-master is running
o hadoop-zookeeper-server
o hadoop-0.20-datanode
o hadoop-0.20-tasktracker
o hbase-regionserver
* 2 Slave nodes running the following services:
o hadoop-0.20-datanode
o hadoop-0.20-tasktracker
o hbase-regionserver
* Running both hbase-master and hbase-regionserver on the Master node.
* 2 1TB storage is attached to the Master node (1 for name dir, 1 for
data dir).
* 1 TB storage is attached to each Slave node (for data dir).
java version "1.6.0_30"
Java(TM) SE Runtime Environment (build 1.6.0_30-b12)
Java HotSpot(TM) 64-Bit Server VM (build 20.5-b03, mixed mode)
I believe the heap is set to the default size. I don't see an -Xmx set
anywhere.
I also have Ganglia running. What is the best metric to describe loading?
P
On 3/13/12 12:00 PM, Stack wrote:
On Tue, Mar 13, 2012 at 8:34 AM, Peter Wolf<[email protected]> wrote:
Hello all,
My Region Server just crashed under load. Apparently, it ran out of memory.
Disappointing...
Does anyone have wisdom on setting Region Server memory? Any reason not to
max it out?
Any chance of a bit of context? Version of hbase and jvm. Hardware.
What load it was carrying at the time. What the loading was like?
Configuration?
Thanks,
St.Ack
