Your screenshot wasn't attached, this ML more often than not strips them. Also this is the expected behavior, the balancer in the master doesn't run all the time... only every 5 min. It really shows up when you start with a fresh table, but you should never import into a single region table anyway since this is slow.
More info: http://hbase.apache.org/book.html#disable.splitting J-D On Wed, Jun 15, 2011 at 12:53 AM, Sujee Maniyam <su...@sujee.net> wrote: > Hi all > I am loading data into a table using a standalone java client (multiple > threads). I see the table splits into regions, but for a while the new > regions are not distributed among region servers. (see screenshot, in this > case all regions are in SN8 machine). So the updates go to one RS as > opposed to many RSs. > Eventually the regions do get spread around. > In RegionServer logs I do see messages like following: > 2011-06-14 15:50:44,319 INFO > org.apache.hadoop.hbase.regionserver.CompactSplitThread: Region split, ME > TA updated, and report to master. > Parent=logs,,1308091368394.a4e9b514dbb9061cd444d5a1104de5b9., new re > gions: logs,,1308091842570.87cc920ee45add9c094c9e872774e2d1., > logs,\x00\x07\xA0\xBC\x00\x00\x00\x01\x7 > F\xFF\xFE\xDA\x17\x03\x1Ac,1308091842570.39dc0057e0bdbcdaa8bd51b5da0dfd3f.. > Split took 1sec > version : Hbase 0.90.1-cdh3u0, r > region size : 1G > is this the expected behavior? > thanks > Sujee > http://sujee.net >