--- Begin Message ---
- From: Devaraj Das <ddas@xxxxxxxxxxxxx>
- To: Yi Kai Tsai <yikai@xxxxxxxxxxxxx>, Vivek Ratan <vivekr@xxxxxxxxxxxxx>
- Date: Tue, 16 Dec 2008 18:51:11 +0900
Hi Yi-Kai, no that's not concrete as yet. People are thinking of various
failover strategies ranging from cold-cold -> warm-warm -> hot-hot. This is
a work in progress currently (and is still under investigation where
potential solutions are Linux HA suite, Zookeeper-based solution, etc.)..
Thanks,
Devaraj
On 12/16/08 3:00 PM, "Yi-Kai Tsai" <yikai@xxxxxxxxxxxxx> wrote:
> hi Devaraj and Vivek
>
> I am looking for further information regarding namenode automatically
> fail-over?
> Its rely on zookeeper right? any JIRA # or roadmap information?
>
> Weeks ago when I gave talk of hadoop, some audience (none yahoo) are
> using hadoop as production component directly,
> though current namenode is not really unstable , but single point
> failure is always a concern.
>
> thanks
--- End Message ---
