Re: fsimage save file

Are you looking at v20 or v23? Note that HDFS-1070 changed the format. But I would describe both the old and the new format as "breadth-first within each directory, depth-first between directories", because when a directory is serialized, all it

Re: fsimage save file

thanks for confirmation. On Mon, May 2, 2011 at 8:48 PM, Todd Lipcon wrote: > On Mon, May 2, 2011 at 6:39 PM, Thanh Do wrote: > >> So the order of saving is based on how child list >> is organized. According from the code from >> INodeDirectory.addChild(), every time we add a child, >> we do a

Re: fsimage save file

On Mon, May 2, 2011 at 6:39 PM, Thanh Do wrote: > So the order of saving is based on how child list > is organized. According from the code from > INodeDirectory.addChild(), every time we add a child, > we do a binary search to find the appropriate insert location. > > So I think the child list i

Re: fsimage save file

Thanks Todd, I looked at FSImage.saveImage(). It turns out the order is depth first. I saw this loop: *for (INode child : current.getChildren()) { ... saveInode2Image(..) } * So the order of saving is based on how child list is organized. According from the code from INodeDirectory.addChild

Re: fsimage save file

On Mon, May 2, 2011 at 1:20 PM, Thanh Do wrote: > hi all, > > perhaps this is a dummy question but > can anyone tell me that when > the namenode saves a fsimage, > are the Inodes saved in an alphabetical order? > Hi Thanh, They're saved in directory traversal order (can't remember if it's bread

fsimage save file

hi all, perhaps this is a dummy question but can anyone tell me that when the namenode saves a fsimage, are the Inodes saved in an alphabetical order? thanks Thanh

Re: Configuration for small Cluster

any comments??? 2011/4/28 baran cakici > Hi Everyone, > > I have a Cluster with one Master(JobTracker and NameNode - Intel Core2Duo 2 > GB Ram) and four Slaves(Datanode and Tasktracker - Celeron 2 GB Ram). My > Inputdata are between 2GB-10GB and I read Inputdata in MapReduce line by > line. Now,