You should be able to modify the dfs.data.dir property and bounce the datanode process.
On 1/21/10 9:30 AM, "Arv Mistry" <[email protected]> wrote: > I had a typo in my original email, which I've corrected below. > But assuming I wanted to add another disk to my slave, should I be able > to do that without losing my current data? Anyone have any documentation > or link they could send me that describes this. > > Appreciate your help, > > Cheers Arv > > -----Original Message----- > From: Arv Mistry [mailto:[email protected]] > Sent: January 21, 2010 11:25 AM > To: [email protected]; [email protected] > Subject: HDFS data storage across multiple disks on slave > > Hi, > > I'm using hadoop 0.20 and trying to understand how hadoop stores the > data. > The setup I have is a single slave, with two disks, 500G each. In the > hdfs-site.xml file I specify for the dfs.data.dir the two disks i.e. > /opt/dfs/data,/opt1/dfs/data. > > Now, a couple of things when I do a report i.e. ./hadoop dfsadmin > -report, > It only says I have a configured capacity of (500G), should that not be > twice that, since there are 2 500G disks. > > And when I look at the data being written, its only written to > /opt/dfs/data. There is no directory /opt1/dfs/data. Should that not > have been created when I formatted the hdfs? > > Could anyone tell me is there an easy way to add this second disk to the > HDFS and preserve the existing data. And any ideas what I did wrong that > it didn't get created/used. > > Any insight would be appreciated. > > Cheers Arv
