I'm thinking xciever limit... Try 6+ machines...
On May 30, 2009 10:37 PM, "Xinan Wu" <[email protected]> wrote: yes epoll and ulimit -n limits are both upped as suggested by docs. My "test cluster" is pretty typical local dev box, ubuntu 8.10 64bit, sun 1.6, 4G mem, hdfs namenode/datanode hbase master/regionserver all running on the same box. My htable has about 10m rows imported when i hit this state... info pages/logs show all regions are ready but I can't pull any data out from them. On Sat, May 30, 2009 at 7:48 AM, stack <[email protected]> wrote: > Tell us more about your cluster ...
