Yes It is case balancer assigned a portion of the total. How many regions in the cluster? Do you say 1344 above? How do we get to 5041?
In my test cluster: 1 hmasters , 2 regionservers , 3 zookeeper and 5041 regions In this scenario: 1, Two Zookeeper crashed 2, One Hmaster and one regionserver crashed 3. zookeeper started 4.Hmaster and regionserver started. 5.I found that region number is more 1000 than the fact. I read the code and found regions should be opened two times. I think region server should be added to onlineServers in two case: 1. region server startup 2. Region server checkin when Hmaster thread call waitForRegionServers() Other case region server should shutdown itself. -----邮件原件----- 发件人: [email protected] [mailto:[email protected]] 代表 Stack 发送时间: 2011年5月25日 12:36 收件人: [email protected] 主题: Re: About RegionServer checkin On Tue, May 24, 2011 at 9:27 PM, Gaojinchao <[email protected]> wrote: > If Region server checkin after call this.serverManager.waitForRegionServers(). > It seems that regionServerReport shouldn't add Region server to onlineServers. > Otherwise The region may be opened again. > I don't follow? Usually what happens is that the next time the balancer runs, this late server gets assigned a portion of the total regions. > 2011-05-23 10:56:33,875 INFO > org.apache.hadoop.hbase.master.AssignmentManager: Bulk assigning 5041 > region(s) across 1 server(s), retainAssignment=true > 2011-05-23 10:56:33,876 DEBUG > org.apache.hadoop.hbase.master.AssignmentManager: Timeout-on-RIT=5041000 How many regions in the cluster? Do you say 1344 above? How do we get to 5041? St.Ack
