Kristoffer, It looks like this is described in https://issues.apache.org/jira/browse/HBASE-8912. In short, stop HBase, remove the affected znode, restart HBase, and then run hbase hbck -fixAssignments . Longer term, it may be worth upgrading past 0.94.16, if possible, because this issue has since been resolved.
Let us know how it goes! -Dima On Tue, Oct 14, 2014 at 1:11 AM, Kristoffer Sjögren <[email protected]> wrote: > I was thinking of doing that but I suspect that zookeeper keeps metadata of > tables also. Seems like region servers are fine for now without the master > and I don't want to make the problem worse by taking chances. > > If I run 'hbase hbck' it complains about master is not running also. > > I found the following two posts which indicate that I should remove the > zookeeper nodes. Can somebody confirm? > > 1. http://stackoverflow.com/a/14557937/1195960 > 2. http://arnon.me/2013/01/killing-hbase-zombie-table/ > > > On Tue, Oct 14, 2014 at 9:51 AM, lagend <[email protected]> wrote: > > > Maybe you can remove the hdfs files of those bad tables. > > like： bin/hadoop fs -rm /hbase/TABLENAME > > > > > > > > > > ------------------ 原始邮件 ------------------ > > 发件人: "Kristoffer Sjögren";<[email protected]>; > > 发送时间: 2014年10月14日(星期二) 下午3:27 > > 收件人: "user"<[email protected]>; > > > > 主题: Force remove table > > > > > > > > Hi > > > > I accidentally created a few tables with 'lzo' compression without the > jar > > dependencies in hbase 0.94.7. This caused havoc on the master which > > eventually crashed and now I can't start it again [1]. > > > > I tried to remove the tables but I can't - hbase shell is complaining > about > > the master is not running so i'm stuck. > > > > Is there a way to forcefully delete these tables or any other way to get > > the master back up? > > > > Cheers, > > -Kristoffer > > > > > > 1. http://pastebin.com/Jprr6j39 > > >
