MongoDB误删表恢复

一、场景描述

    公司某工程师执行db.giveget_card.drop()，误将线上表删除。

    幸好每天都有做备份，这个时候就体现了备份的重要性了，哈哈哈。。。

二、模拟故障过程

备份数据大小：

rs_test01:PRIMARY> use ycsb switched to db ycsb rs_test01:PRIMARY> db.giveget_card.count(); 3173391

删除之前，此表有更新。

rs_test01:PRIMARY> db.giveget_card.insert({id:1}); WriteResult({ "nInserted" : 1 }) rs_test01:PRIMARY> db.giveget_card.insert({id:2}); WriteResult({ "nInserted" : 1 }) rs_test01:PRIMARY> db.giveget_card.insert({id:3}); WriteResult({ "nInserted" : 1 }) rs_test01:PRIMARY> db.giveget_card.insert({id:4}); WriteResult({ "nInserted" : 1 })

其他表也有更新操作。

rs_test01:PRIMARY> db.tab.find(); { "_id" : ObjectId("59354ba202d9a99ab2f879c6"), "name" : "a" } { "_id" : ObjectId("59354ba602d9a99ab2f879c7"), "name" : "b" } { "_id" : ObjectId("59354ba802d9a99ab2f879c8"), "name" : "c" } { "_id" : ObjectId("59354baa02d9a99ab2f879c9"), "name" : "d" }

删除操作之后，此表和其他表都有更新。

rs_test01:PRIMARY> db.giveget_card.find(); { "_id" : ObjectId("59354c28d905432aeaccd53c"), "id" : 5 } { "_id" : ObjectId("59354c2bd905432aeaccd53d"), "id" : 6 } rs_test01:PRIMARY> db.tab.find(); { "_id" : ObjectId("59354ba202d9a99ab2f879c6"), "name" : "a" } { "_id" : ObjectId("59354ba602d9a99ab2f879c7"), "name" : "b" } { "_id" : ObjectId("59354ba802d9a99ab2f879c8"), "name" : "c" } { "_id" : ObjectId("59354baa02d9a99ab2f879c9"), "name" : "d" } { "_id" : ObjectId("59354ccfd905432aeaccd542"), "name" : "e" } { "_id" : ObjectId("59354cd2d905432aeaccd543"), "name" : "f" }

三、恢复步骤

1、将备份中 tab 表的 giveget_card.bson 及 giveget_card.metadata.json 文件拷贝到 /tmp/restore/ycsb 目录（自建目录)，ycsb 为库名。

# cp /data/backup/rs07/ycsb/giveget_card.* /tmp/restore/ycsb

2、将备份时间之后，误删操作之前的 oplog 导出，用于恢复表

# mongodump --port 2203 -d local -c oplog.rs -q '{"ts" : {$gte : Timestamp(1496664480, 10430), $lte : Timestamp(1496665113, 10430)}}' -o /tmp/oplog

--时间戳 是使用转换工具转换之后的结果。

3、使用 bsondump 查看 oplog 日志，找到 drop 操作的时间戳 1496665069

# bsondump /tmp/oplog/local/oplog.rs.bson  {"ts":{"$timestamp":{"t":1496664760,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"7079172056815894727"},"v":2,"op":"i","ns":"ycsb.giveget_card","o":{"_id":{"$oid":"59354ab8c5308d8c7a9da8b5"},"id":1.0}} {"ts":{"$timestamp":{"t":1496664762,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"-1797107728294067016"},"v":2,"op":"i","ns":"ycsb.giveget_card","o":{"_id":{"$oid":"59354abac5308d8c7a9da8b6"},"id":2.0}} {"ts":{"$timestamp":{"t":1496664765,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"8604646791509150392"},"v":2,"op":"i","ns":"ycsb.giveget_card","o":{"_id":{"$oid":"59354abdc5308d8c7a9da8b7"},"id":3.0}} {"ts":{"$timestamp":{"t":1496664768,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"9018614066505371436"},"v":2,"op":"i","ns":"ycsb.giveget_card","o":{"_id":{"$oid":"59354ac0c5308d8c7a9da8b8"},"id":4.0}} {"ts":{"$timestamp":{"t":1496664994,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"-4471524661347063602"},"v":2,"op":"c","ns":"ycsb.$cmd","o":{"create":"tab"}} {"ts":{"$timestamp":{"t":1496664994,"i":2}},"t":{"$numberLong":"12"},"h":{"$numberLong":"-4215905958456607246"},"v":2,"op":"i","ns":"ycsb.tab","o":{"_id":{"$oid":"59354ba202d9a99ab2f879c6"},"name":"a"}} {"ts":{"$timestamp":{"t":1496664998,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"6170506962401844481"},"v":2,"op":"i","ns":"ycsb.tab","o":{"_id":{"$oid":"59354ba602d9a99ab2f879c7"},"name":"b"}} {"ts":{"$timestamp":{"t":1496665000,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"-8071456063660489895"},"v":2,"op":"i","ns":"ycsb.tab","o":{"_id":{"$oid":"59354ba802d9a99ab2f879c8"},"name":"c"}} {"ts":{"$timestamp":{"t":1496665002,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"4387884836668659146"},"v":2,"op":"i","ns":"ycsb.tab","o":{"_id":{"$oid":"59354baa02d9a99ab2f879c9"},"name":"d"}} {"ts":{"$timestamp":{"t":1496665069,"i":1}},"t":{"$numberLong":"12"},"h":{"$numberLong":"-6913449254950935781"},"v":2,"op":"c","ns":"ycsb.$cmd","o":{"drop":"giveget_card"}} 2017-06-05T20:27:25.552+0800 10 objects found

4、将 oplog 的 bson 文件拷贝到相应目录下

# cp /tmp/oplog/local/oplog.rs.bson /tmp/restore/oplog.bson

此时恢复的目录结构：

# pwd /tmp/restore # ls oplog.bson  ycsb

    

5、至此，所有的准备操作已经做完，恢复数据。

[root@ops-db-test02 restore]# mongorestore --port 2203 --oplogReplay --oplogLimit=1496665069:1 /tmp/restore 2017-06-05T20:36:45.361+0800 building a list of dbs and collections to restore from /tmp/restore dir 2017-06-05T20:36:45.364+0800 reading metadata for ycsb.giveget_card from /tmp/restore/ycsb/giveget_card.metadata.json 2017-06-05T20:36:45.364+0800 restoring ycsb.giveget_card from /tmp/restore/ycsb/giveget_card.bson 2017-06-05T20:36:48.362+0800 [........................]  ycsb.giveget_card  15.4MB/475MB  (3.2%) 2017-06-05T20:36:51.362+0800 [#.......................]  ycsb.giveget_card  31.1MB/475MB  (6.6%) 2017-06-05T20:36:54.362+0800 [##......................]  ycsb.giveget_card  46.6MB/475MB  (9.8%) 2017-06-05T20:36:57.362+0800 [###.....................]  ycsb.giveget_card  62.1MB/475MB  (13.1%) 2017-06-05T20:37:00.362+0800 [###.....................]  ycsb.giveget_card  76.4MB/475MB  (16.1%) 2017-06-05T20:37:03.362+0800 [####....................]  ycsb.giveget_card  90.7MB/475MB  (19.1%) 2017-06-05T20:37:06.362+0800 [#####...................]  ycsb.giveget_card  105MB/475MB  (22.0%) 2017-06-05T20:37:09.362+0800 [######..................]  ycsb.giveget_card  120MB/475MB  (25.2%) 2017-06-05T20:37:12.362+0800 [######..................]  ycsb.giveget_card  133MB/475MB  (28.0%) 2017-06-05T20:37:15.362+0800 [#######.................]  ycsb.giveget_card  146MB/475MB  (30.8%) 2017-06-05T20:37:18.363+0800 [########................]  ycsb.giveget_card  163MB/475MB  (34.3%) 2017-06-05T20:37:21.362+0800 [########................]  ycsb.giveget_card  178MB/475MB  (37.4%) 2017-06-05T20:37:24.362+0800 [#########...............]  ycsb.giveget_card  196MB/475MB  (41.3%) 2017-06-05T20:37:27.362+0800 [##########..............]  ycsb.giveget_card  214MB/475MB  (45.0%) 2017-06-05T20:37:30.362+0800 [###########.............]  ycsb.giveget_card  231MB/475MB  (48.6%) 2017-06-05T20:37:33.362+0800 [############............]  ycsb.giveget_card  245MB/475MB  (51.5%) 2017-06-05T20:37:36.362+0800 [#############...........]  ycsb.giveget_card  261MB/475MB  (54.8%) 2017-06-05T20:37:39.362+0800 [##############..........]  ycsb.giveget_card  279MB/475MB  (58.7%) 2017-06-05T20:37:42.362+0800 [###############.........]  ycsb.giveget_card  297MB/475MB  (62.5%) 2017-06-05T20:37:45.362+0800 [###############.........]  ycsb.giveget_card  312MB/475MB  (65.8%) 2017-06-05T20:37:48.362+0800 [################........]  ycsb.giveget_card  328MB/475MB  (69.0%) 2017-06-05T20:37:51.362+0800 [#################.......]  ycsb.giveget_card  341MB/475MB  (71.8%) 2017-06-05T20:37:54.362+0800 [#################.......]  ycsb.giveget_card  356MB/475MB  (74.9%) 2017-06-05T20:37:57.362+0800 [##################......]  ycsb.giveget_card  373MB/475MB  (78.5%) 2017-06-05T20:38:00.362+0800 [###################.....]  ycsb.giveget_card  388MB/475MB  (81.7%) 2017-06-05T20:38:03.362+0800 [####################....]  ycsb.giveget_card  405MB/475MB  (85.2%) 2017-06-05T20:38:06.362+0800 [#####################...]  ycsb.giveget_card  419MB/475MB  (88.2%) 2017-06-05T20:38:09.362+0800 [#####################...]  ycsb.giveget_card  434MB/475MB  (91.4%) 2017-06-05T20:38:12.362+0800 [######################..]  ycsb.giveget_card  442MB/475MB  (93.1%) 2017-06-05T20:38:15.362+0800 [#######################.]  ycsb.giveget_card  459MB/475MB  (96.6%) 2017-06-05T20:38:18.362+0800 [#######################.]  ycsb.giveget_card  475MB/475MB  (99.9%) 2017-06-05T20:38:18.427+0800 [########################]  ycsb.giveget_card  475MB/475MB  (100.0%) 2017-06-05T20:38:18.427+0800 restoring indexes for collection ycsb.giveget_card from metadata 2017-06-05T20:38:44.680+0800 finished restoring ycsb.giveget_card (3173391 documents) 2017-06-05T20:38:44.680+0800 replaying oplog 2017-06-05T20:38:44.739+0800 done

6、查看恢复的结果

rs_test01:PRIMARY> db.giveget_card.find({id : {$gte : 1 }}); { "_id" : ObjectId("59354cb9d905432aeaccd540"), "id" : 5 } { "_id" : ObjectId("59354cc0d905432aeaccd541"), "id" : 6 } { "_id" : ObjectId("59354ab8c5308d8c7a9da8b5"), "id" : 1 } { "_id" : ObjectId("59354abac5308d8c7a9da8b6"), "id" : 2 } { "_id" : ObjectId("59354abdc5308d8c7a9da8b7"), "id" : 3 } { "_id" : ObjectId("59354ac0c5308d8c7a9da8b8"), "id" : 4 }

数据内容相同，但存储顺序与之前数据的存储顺序不同了。

rs_test01:PRIMARY> db.giveget_card.count(); 3173397

结果 count= 备份表数据 3173391+ 之后的更新数据 6 。

    

7、因为 dump 出来的 oplog 也包含了其他表的操作。查看恢复过程中有没有对其他表产生影响。

rs_test01:PRIMARY> db.tab.find(); { "_id" : ObjectId("59354ba202d9a99ab2f879c6"), "name" : "a" } { "_id" : ObjectId("59354ba602d9a99ab2f879c7"), "name" : "b" } { "_id" : ObjectId("59354ba802d9a99ab2f879c8"), "name" : "c" } { "_id" : ObjectId("59354baa02d9a99ab2f879c9"), "name" : "d" } { "_id" : ObjectId("59354ccfd905432aeaccd542"), "name" : "e" } { "_id" : ObjectId("59354cd2d905432aeaccd543"), "name" : "f" }

--查看 tab 表的数据跟原表数据相同，没有什么影响，说明其他表的日志在空跑。

以上就是备份结合 oplog 的恢复操作。

备份很重要！！！ 备份很重要！！！ 备份很重要！！！重要的事情讲三遍～～～