First off, I'm sorry about you losing data. I thought you recognized
that this would be destructive on your data reading that link you sent
out. I wasn't really advising you from a "saving data" standpoint.
Zookeeper crashing (or even `kill -9`ing) should have no effect on
Hadoop. Did Hadoop come up correctly before you tried to restart
Accumulo? Did you then do the `hadoop namenode -format` and expect to
keep your data? If so, lesson learned?
On 3/27/13 3:18 PM, Aji Janis wrote:
Eric and Josh thanks for all your feedback. We ended up _loosing all
our accumulo data_ because I had to reformat hadoop. Here is in a
nutshell what I did:
1. Stop accumulo
2. Stop hadoop
3. On hadoop master and all datanodes, from dfs.data.dir
(hdfs-site.xml) remove everything under the data folder
4. On hadoop master, from dfs.name.dir (hdfs-site.xml) remove
everything under the name folder
5. As hadoop user, execute.../hadoop/bin/hadoop namenode -format
6. As hadoop user, execute.../hadoop/bin/start-all.sh ==> should
populate data/ and name/ dirs that was erased in steps 3, 4.
7. Initialized Accumulo - as accumulo user, ../accumulo/bin/accumulo
init (I created a new instance)
8. Start accumulo
I was wondering if anyone had suggestions or thoughts on how I could
have solved the original issue of accumulo waiting initialization
without loosing my accumulo data? Is it possible to do so?
