Thank you both for such a quick and helpful reply!
My namenode is not good enough to support everything, that's why I need
to distribute some work to other servers :)
Thank you for the links, I'll take them instead of 0.2.0. I wasn't sure
if code on git is stable.
I'm really happy you've created Blur!
On 4/7/2014 5:52 PM, Aaron McCurry wrote:
Hi Neda,
First off welcome!
On Mon, Apr 7, 2014 at 11:27 AM, Neda Grbic <[email protected]> wrote:
Hi,
It seems Blur is all I need for my project, but I have some trouble
finding the best solution, since I do not know anything about zookeeper as
well.
I have two application servers, one hadoop name node server and four data
nodes.
I was wondering if you could give me an idea where to put blur,
zookeper(s), controller and shard servers.
I presume I would have two blur servers on my application servers; shard
servers on data node servers. But I'm not sure where zookeeper(s) and
controllers should be.
I would run blur shard servers on the same nodes as the hadoop data nodes.
As for the controllers and zk nodes, that is really up to you. I agree
with Garret that assuming that your namenode has enough resources you could
run them there. You could also run a controller beside each shard server
but that may be overkill for your current setup. NOTE: Both the
controllers and shard servers need to be able to talk to both ZooKeeper and
HDFS.
I would suggest that you run code from the apache-blur-0.2 branch. The
version will be 0.2.2 when released.
To check it out and compile yourself:
git clone https://git-wip-us.apache.org/repos/asf/incubator-blur.git
git checkout apache-blur-0.2
mvn install -DskipTests -Dhadoop1
You could also grab the latest build binaries from Jenkins Apache CI server:
https://builds.apache.org/job/Blur-master-jdk6/
Download links:
https://builds.apache.org/job/Blur-master-jdk6/212/org.apache.blur$apache-blur/
Let us know if you have any questions.
Aaron
Thanks,
Neda
