On Sat, Jun 12, 2010 at 8:18 PM, Jeff Zhang <zjf...@gmail.com> wrote:
> So when will the replication happen ? Do you know which part of code > control the replication ? Thanks. > > It's all throughout FSNamesystem and related classes. I wouldn't recommend digging in there unless you have a couple days to kill - it gets quite hairy! :) Replication should happen close to immediately unless there is already a large queue of underreplicated blocks. Replication happens through a priority queue, where the priority is based on how far under-replicated a block is. -Todd > > On Sun, Jun 13, 2010 at 11:10 AM, Eric Sammer <esam...@cloudera.com> > wrote: > > There is no way to force replication other than setrep. Do you have > > fewer nodes than configured replicas? Hadoop will not place two > > replicas on one data node, for instance. > > > > On Sat, Jun 12, 2010 at 12:18 AM, Jeff Zhang <zjf...@gmail.com> wrote: > >> But when I invoke command fsck, it show that one replica is missing. > >> So how can I force the replication ? > >> > >> On Sat, Jun 12, 2010 at 12:13 PM, Eric Sammer <esam...@cloudera.com> > wrote: > >>> Jeff: > >>> > >>> The name node will instruct the data nodes to do either remove or > >>> create new replicas. There's nothing additional you need to do other > >>> than fs -setrep. > >>> > >>> On Fri, Jun 11, 2010 at 3:18 AM, Jeff Zhang <zjf...@gmail.com> wrote: > >>>> Hi all, > >>>> > >>>> The command bin/hadoop setrepr only affect the metadata in namenode. > >>>> So How can I make it take effect on data node ? Thanks. > >>>> > >>>> -- > >>>> Best Regards > >>>> > >>>> Jeff Zhang > >>>> > >>> > >>> > >>> > >>> -- > >>> Eric Sammer > >>> phone: +1-917-287-2675 > >>> twitter: esammer > >>> data: www.cloudera.com > >>> > >> > >> > >> > >> -- > >> Best Regards > >> > >> Jeff Zhang > >> > > > > > > > > -- > > Eric Sammer > > phone: +1-917-287-2675 > > twitter: esammer > > data: www.cloudera.com > > > > > > -- > Best Regards > > Jeff Zhang > -- Todd Lipcon Software Engineer, Cloudera