Scaling _down_ is a continual problem for us, and this is one of the prime factors. It puts a bad taste in the mouth of new people who then run away from HBase and HDFS since it is "unreliable and unstable". It is perfectly within scope to support a cluster of about 5-6 machines which can have an aggregate capacity of 24TB (which is a fair amount), and people expect to start small, prove the concept/technology then move up.
I am also +1 On Thu, Jan 21, 2010 at 2:36 PM, Stack <st...@duboce.net> wrote: > I'd like to propose a new vote on having hdfs-630 committed to 0.21. > The first vote on this topic, initiated 12/14/2009, was sunk by Tsz Wo > (Nicholas), Sze suggested improvements. Those suggestions have since > been folded into a new version of the hdfs-630 patch. Its this new > version of the patch -- 0001-Fix-HDFS-630-0.21-svn-2.patch -- that I'd > like us to vote on. For background on why we -- the hbase community > -- think hdfs-630 important, see the notes below from the original > call-to-vote. > > I'm obviously +1. > > Thanks for you consideration, > St.Ack > > P.S. Regards TRUNK, after chatting with Nicholas, TRUNK was cleaned of > the previous versions of hdfs-630 and we'll likely apply > 0001-Fix-HDFS-630-trunk-svn-4.patch, a version of > 0001-Fix-HDFS-630-0.21-svn-2.patch that works for TRUNK that includes > the Nicholas suggestions. > > > On Mon, Dec 14, 2009 at 9:56 PM, stack <st...@duboce.net> wrote: >> I'd like to propose a vote on having hdfs-630 committed to 0.21 (Its already >> been committed to TRUNK). >> >> hdfs-630 adds having the dfsclient pass the namenode the name of datanodes >> its determined dead because it got a failed connection when it tried to >> contact it, etc. This is useful in the interval between datanode dying and >> namenode timing out its lease. Without this fix, the namenode can often >> give out the dead datanode as a host for a block. If the cluster is small, >> less than 5 or 6 nodes, then its very likely namenode will give out the dead >> datanode as a block host. >> >> Small clusters are common in hbase, especially when folks are starting out >> or evaluating hbase. They'll start with three or four nodes carrying both >> datanodes+hbase regionservers. They'll experiment killing one of the slaves >> -- datanodes and regionserver -- and watch what happens. What follows is a >> struggling dfsclient trying to create replicas where one of the datanodes >> passed us by the namenode is dead. DFSClient will fail and then go back to >> the namenode again, etc. (See >> https://issues.apache.org/jira/browse/HBASE-1876 for more detailed >> blow-by-blow). HBase operation will be held up during this time and >> eventually a regionserver will shut itself down to protect itself against >> dataloss if we can't successfully write HDFS. >> >> Thanks all, >> St.Ack >
