mongodb之 mongodump 与 mongorestore

一、备份

和之前介绍的 mongoexport 的数据导出工具不同， mongodump 是将数据以二进制形式导出，而 mongoexport 导出的数据格式为 csv 或 json 格式； mongodump 可以导出一个数据库，或者整个 MongoDB 服务上的所有数据库，因此 mongodump 是更大范围的备份工具。

[root@mysql3 admin]# mongodump --help
Usage:
mongodump <options>

Export the content of a running server into .bson files.

Specify a database with -d and a collection with -c to only dump that database or collection.

See http://docs.mongodb.org/manual/reference/program/mongodump/ for more information.

general options:
--help print usage
--version print the tool version and exit

verbosity options:
-v, --verbose=<level> more detailed log output (include multiple times for more verbosity, e.g.
-vvvvv, or specify a numeric value, e.g. --verbose=N)
--quiet hide all log output

connection options:
-h, --host=<hostname> mongodb host to connect to (setname/host1,host2 for replica sets)
--port=<port> server port (can also use --host hostname:port)

ssl options:
--ssl connect to a mongod or mongos that has ssl enabled
--sslCAFile=<filename> the .pem file containing the root certificate chain from the certificate
authority
--sslPEMKeyFile=<filename> the .pem file containing the certificate and key
--sslPEMKeyPassword=<password> the password to decrypt the sslPEMKeyFile, if necessary
--sslCRLFile=<filename> the .pem file containing the certificate revocation list
--sslAllowInvalidCertificates bypass the validation for server certificates
--sslAllowInvalidHostnames bypass the validation for server name
--sslFIPSMode use FIPS mode of the installed openssl library

authentication options:
-u, --username=<username> username for authentication
-p, --password=<password> password for authentication
--authenticationDatabase=<database-name> database that holds the user's credentials
--authenticationMechanism=<mechanism> authentication mechanism to use

namespace options:
-d, --db=<database-name> database to use
-c, --collection=<collection-name> collection to use

query options:
-q, --query= query filter, as a JSON string, e.g., '{x:{$gt:1}}'
--queryFile= path to a file containing a query filter (JSON)
--readPreference=<string>|<json> specify either a preference name or a preference json object
--forceTableScan force a table scan

output options:
-o, --out=<directory-path> output directory, or '-' for stdout (defaults to 'dump')
--gzip compress archive our collection output with Gzip
--repair try to recover documents from damaged data files (not supported by all
storage engines)
--oplog use oplog for taking a point-in-time snapshot
--archive=<file-path> dump as an archive to the specified path. If flag is specified without a
value, archive is written to stdout
--dumpDbUsersAndRoles dump user and role definitions for the specified database
--excludeCollection=<collection-name> collection to exclude from the dump (may be specified multiple times to
exclude additional collections)
--excludeCollectionsWithPrefix=<collection-prefix> exclude all collections from the dump that have the given prefix (may be
specified multiple times to exclude additional prefixes)
-j, --numParallelCollections= number of collections to dump in parallel (4 by default) (default: 4)
--viewsAsCollections dump views as normal collections with their produced data, omitting
standard collections

# 备份指定数据库中的某个集合
mongodump -d andy -h 10.100.25.42:27001 -o mongo_andy.dmp -c XX

# 备份指定数据库
mongodump -d andy -h 10.100.25.42 --port 27001 -o mongo_andy.dmp

说明:从日志看出，mongodump 的结果会生成一个目录，第一层目录为数据库名，再下一层为每个集合对应的备份文件。

# 备份实例所有数据库
mongodump -h 10.100.25.42 --port 27001 -o mongo_all.dmp

说明：mongodump 如果不指定 -d 参数，刚会备份整个 MongoDB 实例。

二、恢复

[root@mysql3 admin]# mongorestore --help
Usage:
mongorestore <options> <directory or file to restore>

Restore backups generated with mongodump to a running server.

Specify a database with -d to restore a single database from the target directory,
or use -d and -c to restore a single collection from a single .bson file.

See http://docs.mongodb.org/manual/reference/program/mongorestore/ for more information.

general options:
--help print usage
--version print the tool version and exit

verbosity options:
-v, --verbose=<level> more detailed log output (include multiple times for more verbosity, e.g.
-vvvvv, or specify a numeric value, e.g. --verbose=N)
--quiet hide all log output

connection options:
-h, --host=<hostname> mongodb host to connect to (setname/host1,host2 for replica sets)
--port=<port> server port (can also use --host hostname:port)

ssl options:
--ssl connect to a mongod or mongos that has ssl enabled
--sslCAFile=<filename> the .pem file containing the root certificate chain from the certificate
authority
--sslPEMKeyFile=<filename> the .pem file containing the certificate and key
--sslPEMKeyPassword=<password> the password to decrypt the sslPEMKeyFile, if necessary
--sslCRLFile=<filename> the .pem file containing the certificate revocation list
--sslAllowInvalidCertificates bypass the validation for server certificates
--sslAllowInvalidHostnames bypass the validation for server name
--sslFIPSMode use FIPS mode of the installed openssl library

authentication options:
-u, --username=<username> username for authentication
-p, --password=<password> password for authentication
--authenticationDatabase=<database-name> database that holds the user's credentials
--authenticationMechanism=<mechanism> authentication mechanism to use

namespace options:
-d, --db=<database-name> database to use when restoring from a BSON file
-c, --collection=<collection-name> collection to use when restoring from a BSON file
--excludeCollection=<collection-name> DEPRECATED; collection to skip over during restore (may be specified
multiple times to exclude additional collections)
--excludeCollectionsWithPrefix=<collection-prefix> DEPRECATED; collections to skip over during restore that have the given
prefix (may be specified multiple times to exclude additional prefixes)
--nsExclude=<namespace-pattern> exclude matching namespaces
--nsInclude=<namespace-pattern> include matching namespaces
--nsFrom=<namespace-pattern> rename matching namespaces, must have matching nsTo
--nsTo=<namespace-pattern> rename matched namespaces, must have matching nsFrom

input options:
--objcheck validate all objects before inserting
--oplogReplay replay oplog for point-in-time restore
--oplogLimit=<seconds>[:ordinal] only include oplog entries before the provided Timestamp
--oplogFile=<filename> oplog file to use for replay of oplog
--archive=<filename> restore dump from the specified archive file. If flag is specified
without a value, archive is read from stdin
--restoreDbUsersAndRoles restore user and role definitions for the given database
--dir=<directory-name> input directory, use '-' for stdin
--gzip decompress gzipped input

restore options:
--drop drop each collection before import
--dryRun view summary without importing anything. recommended with verbosity
--writeConcern=<write-concern> write concern options e.g. --writeConcern majority, --writeConcern '{w:
3, wtimeout: 500, fsync: true, j: true}' (defaults to 'majority')
(default: majority)
--noIndexRestore don't restore indexes
--noOptionsRestore don't restore collection options
--keepIndexVersion don't update index version
--maintainInsertionOrder preserve order of documents during restoration
-j, --numParallelCollections= number of collections to restore in parallel (4 by default) (default: 4)
--numInsertionWorkersPerCollection= number of insert operations to run concurrently per collection (1 by
default) (default: 1)
--stopOnError stop restoring if an error is encountered on insert (off by default)
--bypassDocumentValidation bypass document validation

# 追加恢复
mongorestore -h 10.100.25.42:27001 /tmp/mongo_andy.dmp

# 恢复前先删除数据库中与备份集相同对象
mongorestore -h 10.100.25.42:27001 --drop /tmp/mongo_andy.dmp