解决GREENPLUM某些版本gprecoverseg –r失败后镜像双坏，系统无法启动的问题
?開始之前，我先聲明一點(diǎn)，參考這篇文章去操作數(shù)據(jù)庫，造成丟失數(shù)據(jù)，搞癱系統(tǒng)的可能性非常之高，初學(xué)者以及思路不清晰者，請關(guān)閉頁面，不要再繼續(xù)下去，經(jīng)過一通亂搞以后，真的就沒人能再幫得了你了。
先說明一下故障的來由，由于系統(tǒng)的BUG某些版本的GREENPLUM只要執(zhí)行了gprecoverseg -r，就一定會失敗，而且一定會鏡像雙壞，無法啟動。這里說明一下，這種版本的GP鏡像反轉(zhuǎn)時，只能重啟數(shù)據(jù)庫去自動修復(fù)，不可以使用gprecoverseg –r人工修復(fù)，切記。
????值得慶幸的是，只要是進(jìn)入到這種反轉(zhuǎn)修復(fù)失敗的情況，一般來說所有鏡像一定是同步的，因?yàn)槿绻煌降脑挄?bào)另一個錯誤：有部分鏡像失敗，無法執(zhí)行g(shù)precoverseg -r，這種情況就不會出現(xiàn)我們本文討論的問題。此時雖然我們無法啟動整個數(shù)據(jù)庫，但是事實(shí)上MIRROR是可以作為PRIMARY啟動的，只是GREENPLUM把節(jié)點(diǎn)的狀態(tài)搞亂了。我們只要把出現(xiàn)問題之前的描述節(jié)點(diǎn)狀態(tài)的系統(tǒng)表gp_segment_configuration恢復(fù)到鏡像反轉(zhuǎn)的狀態(tài)，就可以啟動數(shù)據(jù)庫了，當(dāng)然只能是手動恢復(fù)。
接下來我結(jié)合實(shí)際的數(shù)據(jù)介紹一下gp_segment_configuration?的表結(jié)構(gòu)?
postgres=# select * from gp_segment_configuration??where address like 'sdw10-%' or address like 'sdw11-%' order by dbid ;
?dbid | content | role | preferred_role | mode | status | port??|??hostname???| address | replication_port | san_mounts
------+---------+------+----------------+------+--------+-------+-------------+---------+------------------+------------
???56 |??????54 | p ????| p??????????????| c????| u??????| 40000 | dssln-sdw10 | sdw10-1 |????????????41000 |
???57 |??????55 | p????| p??????????????| c????| u??????| 40001 | dssln-sdw10 | sdw10-1 |????????????41001 |
???58 |??????56 | p????| p??????????????| c????| u??????| 40002 | dssln-sdw10 | sdw10-1 |????????????41002 |
???59 |??????57 | p????| p??????????????| c????| u??????| 40003 | dssln-sdw10 | sdw10-2 |????????????41003 |
???60 |??????58 | p????| p??????????????| c????| u??????| 40004 | dssln-sdw10 | sdw10-2 |????????????41004 |
???61 |??????59 | p????| p??????????????| c????| u??????| 40005 | dssln-sdw10 | sdw10-2 |????????????41005 |
???62 |??????60 | p????| p??????????????| s????| u??????| 40000 | dssln-sdw11 | sdw11-1 |????????????41000 |
???63 |??????61 | p????| p??????????????| s????| u??????| 40001 | dssln-sdw11 | sdw11-1 |????????????41001 |
???64 |??????62 | p????| p??????????????| s????| u??????| 40002 | dssln-sdw11 | sdw11-1 |????????????41002 |
???65 |??????63 | p????| p??????????????| s????| u??????| 40003 | dssln-sdw11 | sdw11-2 |????????????41003 |
???66 |??????64 | p????| p??????????????| s????| u??????| 40004 | dssln-sdw11 | sdw11-2 |????????????41004 |
???67 |??????65 | p????| p??????????????| s????| u??????| 40005 | dssln-sdw11 | sdw11-2 |????????????41005 |
??158 |??????48 | m????| m??????????????| s????| u??????| 50000 | dssln-sdw10 | sdw10-2 |????????????51000 |
??159 |??????49 | m????| m??????????????| s????| u??????| 50001 | dssln-sdw10 | sdw10-2 |????????????51001 |
??160 |??????50 | m????| m??????????????| s????| u??????| 50002 | dssln-sdw10 | sdw10-2 |????????????51002 |
??161 |??????51 | m????| m??????????????| s????| u??????| 50003 | dssln-sdw10 | sdw10-1 |????????????51003 |
??162 |??????52 | m????| m??????????????| s????| u??????| 50004 | dssln-sdw10 | sdw10-1 |????????????51004 |
??163 |??????53 | m????| m??????????????| s????| u??????| 50005 | dssln-sdw10 | sdw10-1 |????????????51005 |
??164 |??????54 | m????| m??????????????| c????| d??????| 50000 | dssln-sdw11 | sdw11-2 |????????????51000 |
??165 |??????55 | m????| m??????????????| c????| d??????| 50001 | dssln-sdw11 | sdw11-2 |????????????51001 |
??166 |??????56 | m????| m??????????????| c????| d??????| 50002 | dssln-sdw11 | sdw11-2 |????????????51002 |
??167 |??????57 | m????| m??????????????| c????| d??????| 50003 | dssln-sdw11 | sdw11-1 |????????????51003 |
??168 |??????58 | m????| m??????????????| c????| d??????| 50004 | dssln-sdw11 | sdw11-1 |????????????51004 |
??169 |??????59 | m????| m??????????????| c????| d??????| 50005 | dssln-sdw11 | sdw11-1 |???????????51005 |
?dbid | content | role | preferred_role | mode | status | port??|??hostname???| address | replication_port | san_mounts
dbid?和content表示數(shù)據(jù)庫的ID
role?代表當(dāng)前角色
preferred_role?代表首選角色，也就是原本應(yīng)該出演的角色
mode='s/c/r'三個取值分別代表synced, change logging, resyncing
status='u/d'兩個取值分別代表up，down。
其他字段不做解釋了。
上面這個樣例數(shù)據(jù)表示11節(jié)點(diǎn)的6個鏡像子庫全部被標(biāo)記為DOWN了，但是事實(shí)上里邊的數(shù)據(jù)是完整的，10節(jié)點(diǎn)的6個主用子庫鏡像失敗，可以啟動，但是啟動不起來，但事實(shí)上里邊的數(shù)據(jù)也是完整的只是狀態(tài)標(biāo)記出現(xiàn)了問題。
根據(jù)之前的分析，我們只要把這兩個節(jié)點(diǎn)的鏡像數(shù)據(jù)到反轉(zhuǎn)狀態(tài)就可以了。目標(biāo)如下：
postgres=# select * from gp_segment_configuration??where (address like 'sdw10-%' and preferred_role = ‘p’) or (address like 'sdw11-%' ?and preferred_role = ‘m’) order by dbid ;
?dbid | content | role | preferred_role | mode | status | port??|??hostname???| address | replication_port | san_mounts
------+---------+------+----------------+------+--------+-------+-------------+---------+------------------+------------
???56 |??????54 | m ????| p??????????????| s????| u??????| 40000 | dssln-sdw10 | sdw10-1 |????????????41000 |
???57 |??????55 | m????| p??????????????| s ???| u??????| 40001 | dssln-sdw10 | sdw10-1 |????????????41001 |
???58 |??????56 | m????| p??????????????| s????| u??????| 40002 | dssln-sdw10 | sdw10-1 |????????????41002 |
???59 |??????57 | m????| p??????????????| s????| u??????| 40003 | dssln-sdw10 | sdw10-2 |????????????41003 |
???60 |??????58 | m????| p??????????????| s????| u??????| 40004 | dssln-sdw10 | sdw10-2 |????????????41004 |
???61 |??????59 | m????| p??????????????| s????| u??????| 40005 | dssln-sdw10 | sdw10-2 |????????????41005 |
??164 |??????54 | p????| m??????????????| s????| u??????| 50000 | dssln-sdw11 | sdw11-2 |????????????51000 |
??165 |??????55 | p????| m??????????????| s????| u??????| 50001 | dssln-sdw11 | sdw11-2 |????????????51001 |
??166 |??????56 | p????| m??????????????| s????| u??????| 50002 | dssln-sdw11 | sdw11-2 |????????????51002 |
??167 |??????57 | p????| m??????????????| s????| u??????| 50003 | dssln-sdw11 | sdw11-1 |????????????51003 |
??168 |??????58 | p????| m??????????????| s????| u??????| 50004 | dssln-sdw11 | sdw11-1 |????????????51004 |
??169 |??????59 | p????| m??????????????| s????| u??????| 50005 | dssln-sdw11 | sdw11-1 |???????????51005 |
好我們正式開始：
1、???????啟動到維護(hù)模式：
gpstart –m
以維護(hù)方式連接到數(shù)據(jù)庫
PGOPTIONS='-c gp_session_role=utility' psql -d postgres
打開系統(tǒng)表的維護(hù)開關(guān)
set allow_system_table_mods=DML;
2、???????備份系統(tǒng)表:
create table mybak_segment_configuration as select * from gp_segment_configuration;
3、???????修改數(shù)據(jù)到鏡像反轉(zhuǎn)狀態(tài)：
update gp_segment_configuration set role='p',mode='s',status='u'?where address like 'sdw11%' and preferred_role='m';
update gp_segment_configuration set role='m',mode='s',status='u'?where address like 'sdw10%' and preferred_role='p';
\q
4、???????重啟數(shù)據(jù)庫
gpstop –m –M fast
gpstart
5、???????檢查鏡像，此時應(yīng)該已經(jīng)自動恢復(fù)到正常狀態(tài)。
gpstate –m
6、???????如果不同步就修一下
gprecoverseg
7、???????如果鏡像仍然反轉(zhuǎn)就重新啟動一下數(shù)據(jù)庫。
gpstop –M fast
gpstart
分享：解決GREENPLUM某些版本gprecoverseg?–r失敗后鏡像雙壞，系統(tǒng)無法啟動的問題_醉糊涂蟲_新浪博客
總結(jié)
以上是生活随笔為你收集整理的解决GREENPLUM某些版本gprecoverseg –r失败后镜像双坏，系统无法启动的问题的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。
- 上一篇： C++中单独大括号的意义
- 下一篇： java信息管理系统总结_java实现科