On 07/25/2011 12:19 PM, Mahadev Konar wrote:
Hi Prashanth,
Sorry to disappoint :). Thats isnt true.
Folks in HDFS are working on this currently.
Sanjay has uploaded a design doc in case you want to check it out:
https://issues.apache.org/jira/browse/HDFS-1623
thanks
mahadev
On Sun, Jul 24, 2011 at 11:44 PM, Prashant<prashan...@imaginea.com> wrote:
On 07/19/2011 12:35 AM, Yang wrote:
as described in
http://www.google.com/url?sa=t&source=web&cd=1&ved=0CBgQFjAA&url=http%3A%2F%2Fhadoopblog.blogspot.com%2F2010%2F02%2Fhadoop-namenode-high-availability.html&ei=D4MkTsCrI8TRiALN4-mvAw&usg=AFQjCNFTIAn6ETjAxlC6s5u33sjTUARZQw&sig2=yIka9cpvn9qHvve95niefA
and in the recent FB paper, AvatarNode is a different solution as the
BackUpNode/CheckpointNode feature, right?
-----since the link above says: "The HDFS BackupNode is not available
in 0.20, and upgrading our cluster to hadoop 0.20 is not a feasible
option for us because .... ".
if they are different, is the FB AvatarNode feature is open-sourced ?
if so, is it a better or worse solution than the current
BackUpNode/CheckpointNode in apache distribution?
Thanks
yang
I think in current trunk version MR-279 branch hadoop has incorporated HA
for namenode using the Zookeeper service. I think that is a better technique
than Avatarnode's. I hope somebody can through light on this.
Thats ok!, Actually I confused it with HA for jobtracker which is there
in MR-279.
Thanks for correcting.
