http://abloz.com date:2012.8.14

上一篇文章《hbase 複製備份資料》 中提到用工具CopyTable來在叢集間複製資料。另外還有一種更暴力的方式，來共享HBase備份表。因為有時候兩個叢集並不連通。

一、從源hbase叢集中複製出HBase資料庫表到本地目錄

最好停止HBase，否則可能會丟部分資料

[[email protected] ~]$ hadoop fs -get /hbase/toplist_ware_total_1009_201232 toplist_ware_total_1009_201232

壓縮

[[email protected]
 
 ~]$ tar zcvf topl.tar.gz toplist_ware_total_1009_201232

遠端複製到目標機器

[[email protected] ~]$ scp topl.tar.gz [email protected]:~/.

二、目標HBase匯入

解壓 [[email protected] ~]$ tar zxvf topl.tar.gz

如果目標HBase裡有這個表，需disable並drop掉。如果有該目錄，則用hadoop fs -rmr /hbase/table的方式刪除，再往HDFS上覆制。以免資料出錯。

放到叢集下面

[
 
[email protected] ~]$ fs -put toplist_ware_total_1009_201232 /hbase
[[email protected] ~]$

此時可以list出來，但scan報錯

hbase(main):055:0> list 'toplist_ware_total_1009_201232'
TABLE
toplist_ware_total_1009_201232
1 row(s) in 0.0220 seconds

hbase(main):062:0> scan 'toplist_ware_total_1009_201232'
ROW                                        COLUMN+CELL

ERROR: Unknown table toplist_ware_total_1009_201232!
 

.META.表裡面沒有相關記錄 hbase(main):064:0> scan ‘.META.’ 裡面沒有toplist_ware_total_1009_201232 開頭的行

三、修復.META.表和重新分配資料到各RegionServer

在.META.表沒修復時執行重新分配，會報錯

[[email protected] ~]$ hbase hbck -fixAssignments
...
ERROR: Region { meta => null, hdfs => hdfs://h185:54310/hbase/toplist_ware_total_1009_201232/0403552001eb2a31990e443dcae74ee8, deployed =>  } on HDFS, but not listed in META or deployed on any region server
...

先修復.META.表

[[email protected] ~]$ hbase hbck -fixMeta
...
ERROR: Region { meta => null, hdfs => hdfs://h185:54310/hbase/toplist_ware_total_1009_201232/0403552001eb2a31990e443dcae74ee8, deployed =>  } on HDFS, but not listed in META or deployed on any region server
12/08/14 18:25:15 INFO util.HBaseFsck: Patching .META. with .regioninfo: {NAME => 'toplist_ware_total_1009_201232,,1344187094829.0403552001eb2a31990e443dcae74ee8.', STARTKEY => '', ENDKEY => '', ENCODED => 0403552001eb2a31990e443dcae74ee8,}
...

此時.META.表已經有表的資料了，但scan還是失敗
hbase(main):065:0> scan '.META.'
ROW                                        COLUMN+CELL
...
toplist_ware_total_1009_201232,,134418709 column=info:regioninfo, timestamp=1344939930752, value={NAME => 'toplist_ware_total_1009_201232,,1344187094829.0403552001eb
 4829.0403552001eb2a31990e443dcae74ee8.    2a31990e443dcae74ee8.', STARTKEY => '', ENDKEY => '', ENCODED => 0403552001eb2a31990e443dcae74ee8,}
16 row(s) in 0.0550 seconds

scan還是失敗
hbase(main):066:0> scan 'toplist_ware_total_1009_201232'
ROW                                        COLUMN+CELL

ERROR: org.apache.hadoop.hbase.client.NoServerForRegionException: No server address listed in .META. for region toplist_ware_total_1009_201232,,1344187094829.0403552001eb2a31990e443dcae74ee8. containing row

重新分配到各分割槽伺服器

[[email protected] ~]$ hbase hbck -fixAssignments
...
ERROR: Region { meta => toplist_ware_total_1009_201232,,1344187094829.0403552001eb2a31990e443dcae74ee8., hdfs => hdfs://h185:54310/hbase/toplist_ware_total_1009_201232/0403552001eb2a31990e443dcae74ee8, deployed =>  } not deployed on any region server.
Trying to fix unassigned region...
12/08/14 18:28:01 INFO util.HBaseFsckRepair: Region still in transition, waiting for it to become assigned: {NAME => 'toplist_ware_total_1009_201232,,1344187094829.0403552001eb2a31990e443dcae74ee8.', STARTKEY => '', ENDKEY => '', ENCODED => 0403552001eb2a31990e443dcae74ee8,}
12/08/14 18:28:02 INFO util.HBaseFsckRepair: Region still in transition, waiting for it to become assigned: {NAME => 'toplist_ware_total_1009_201232,,1344187094829.0403552001eb2a31990e443dcae74ee8.', STARTKEY => '', ENDKEY => '', ENCODED => 0403552001eb2a31990e443dcae74ee8,}
12/08/14 18:28:04 INFO util.HBaseFsckRepair: Region still in transition, waiting for it to become assigned: {NAME => 'toplist_ware_total_1009_201232,,1344187094829.0403552001eb2a31990e443dcae74ee8.', STARTKEY => '', ENDKEY => '', ENCODED => 0403552001eb2a31990e443dcae74ee8,}
12/08/14 18:28:05 INFO util.HBaseFsckRepair: Region still in transition, waiting for it to become assigned: {NAME => 'toplist_ware_total_1009_201232,,1344187094829.0403552001eb2a31990e443dcae74ee8.', STARTKEY => '', ENDKEY => '', ENCODED => 0403552001eb2a31990e443dcae74ee8,}
...

scan成功!

hbase(main):067:0> scan 'toplist_ware_total_1009_201232'
ROW                                        COLUMN+CELL
 0000000001                                column=info:loginid, timestamp=1344187147972, value=jjm167258611
 0000000001                                column=info:nick, timestamp=1344187147972, value=?xE9x97xB4?xE6xB5xA3?
 0000000001                                column=info:score, timestamp=1344187147972, value=200
 0000000001                                column=info:userid, timestamp=1344187147972, value=167258611
...
330 row(s) in 0.8630 seconds

如果目標叢集是空的，則可以直接將源HBase的/hbase目錄複製出來，然後在目標HBase系統上fs -rmr /hbase 或fs -mv /hbase /hbase1 然後用hadoop fs -put hbase / 即可

如非註明轉載, 均為原創. 本站遵循知識共享CC協議,轉載請註明來源