Not sure this is the case here because EC2 startup scripts start master node
first and only then start the rest.

Alexander

2008/11/20 Steve Loughran <[EMAIL PROTECTED]>

> Alexander Aristov wrote:
>
>
>>>> What can be a problem that hadoop doesn't add datanodes on startup but
>>>> adds
>>>> them after rebooting.
>>>>
>>>
> Could be a bit that recent hadoop versions may not retry reliably if the
> datanode/task tracker comes up before the "master" node. If the namenode is
> taking longer to come up than the datanodes, it may not be binding properly.
>



-- 
Best Regards
Alexander Aristov

Reply via email to