hadoop-common-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Harsh J <ha...@cloudera.com>
Subject Re: refresh namenode topology cache
Date Tue, 31 Jan 2012 05:42:09 GMT

On Tue, Jan 31, 2012 at 12:24 AM, Sateesh Lakkarsu <lakkarsu@gmail.com> wrote:
> A couple of nodes got registered under /default-rack as NN was unable to
> resolve when the nodes got added. NN can resolve them now, so wondering how
> to fix this without bouncing NN... any way to have it refresh cache?
> Thanks.

Adding completely new nodes should not be a problem if you updated
your rack configuration script/mapping file first, and then started
the service. This would mean that since there's no element in cache,
the lookup happens via the provided script and hence resolves

The default script based mapping implementation does not provide a
mapping refresh option presently, but the implementation can be
over-ridden to use your own class in case you want to implement one
which can be commanded to flush the cache.

Harsh J
Customer Ops. Engineer, Cloudera

View raw message