Skip to content

Latest commit

 

History

History
251 lines (195 loc) · 11.5 KB

README.md

File metadata and controls

251 lines (195 loc) · 11.5 KB
Raw

riak-data-migrator

Tool for migrating data from one or more buckets in a Riak K/V store into another Riak cluster by exporting all data from buckets to disk, and then allowing the user to load one or more of the dumped buckets back into another Riak K/V host or cluster.

Intended Usage

This data migrator tool may be helpful in the following scenarios:

  1. Migrating an entire non-live Riak cluster (no new writes are coming in) to a cluster with a different ring size.
  2. Exporting the contents of a single bucket to disk (again, on a non-live cluster where no new writes are coming in)
  3. Exporting a list of keys only, from a particular bucket to a text file (same conditions as above)
  4. Taking an approximate snapshot of live data for a QA cluster (approximate because, if new objects are written to the cluster after the export operation starts, they are not guaranteed to be exported in that session)

The app works by performing a streaming List Keys operation on one or more Riak buckets, and then issuing GETs for the resulting keys (parallelized as much as possible) and storing the Riak objects (in Protocol Buffer format) on disk. The key listing alone involves multiple iterations through the entire Riak keyspace, and is not intended for frequent usage on a live cluster. On the import side, the app reads the exported Riak objects from files on disk, and issues PUTs to the target cluster.

Do NOT use for regular data backup on a live cluster. Instead, see the Backing Up Riak documentation for recommended best practices. Again, the reason for this admonition is: if new data is written to a live cluster after the export operation is started, that new data is not guaranteed to be included in the export.

Workflow

To transfer data from one Riak cluster to another:

  1. Before using the migrator tool, make sure that the app.config files are the same in both clusters. Meaning, settings such as default quorum values, multi-backend and MDC replication settings, should be the same before starting export/import operations.
  2. (Optional) If applicable, transfer custom bucket properties using the -d -t options to export from one cluster and -l -t to import into the target cluster. Note: Do this only if you set custom properties such as non-default quorum values, pre- or post-commit hooks, or set up Search indexing on a bucket.
  3. Export the contents of a bucket (Riak objects) using the -d option, to files on disk (the objects will be stored in the binary ProtoBuf format)
  4. (Optional, Search-only) If backing up Search-indexed buckets using Data Migrator versions <= 0.2.6, go into the exported data directory and delete the internal-use-only indexing buckets (rm -rf _rsid_*). See #4 for explanation.
  5. Load the Riak objects from the exported files into the target cluster using the -l option.

Downloading:

You can download the ready to run jar file at: https://s3-us-west-2.amazonaws.com/ps-tools/riak-data-migrator-0.2.9-bin.tar.gz

After downloading, unzip/untar it, and it's ready to run from its directory.

tar -xvzf riak-data-migrator-0.2.9-bin.tar.gz
cd riak-data-migrator-0.2.9
java -jar riak-data-migrator-0.2.9.jar [options]

Building from source:

  1. Make sure Apache Maven is installed

  2. Build the Riak Data Migrator itself, using Maven. First, fork this project, and git clone your fork.

cd riak-data-migrator
mvn clean
mvn package
The compiled .jar file is located in the ```target/``` directory.
The usable binary file is ```riak-data-migrator-0.2.9-bin.tar.gz```

Usage:

Usage:

java -jar riak-data-migrator-0.2.9.jar [options]

Options:

Data Transfer (required, one of: d, l, k, or delete)
-d Export (Dump) the contents bucket(s) (keys and objects), in ProtoBuf format, to files
-l Import (Load) the contents of bucket(s) exported by the data migrator, from files.
-k Export a list of Keys only (not object values) from bucket(s), to a text file. (Not to be used with -t)

Settings Transfer (optional, used with to -d or -l)
-t Transfer custom bucket properties only (no data), to and from files on disk. 
   You must specify one or more buckets (not to be used with -a, the All Buckets option).
   You must also specify -d to export or -l to import, with this option.

Delete a bucket
--delete Delete bucket data. Cannot be used with -d, -l, -k, or -t. Must be used with -b or -f

Path (required)
-r <path> Set the path for data to be loaded to or dumped from (path must be valid)

Bucket Options (required for -d, -k or -t)
-a Export all buckets.
-b <bucket name> Export a single bucket.  
-f <bucketNameFile.txt> Export multiple buckets listed in a file (containing line-delimited bucket names)
--loadkeys <bucketKeyNameFile.txt> Specify multiple keys listed in a file (containing line-delimited bucket,key names). Not functional with the -l option.
--bucketkeys <keysFiles.txt> Specify keys listed in a line-delimited file in conjunction with the -b option to specify which bucket the keys belong to. Not functional with the -l option.

Cluster Addresses and Ports (required)
-h <hostName> Specify Riak hostname. Required if a cluster host name file is not specified.  
-c <clusterNameFile.txt> Specify a file containing line delimited Riak Cluster Host
   Names. Required if a host name is not specified. Host name is not specified.  
-p <pbcPortNumber> Specify Riak Protocol Buffers Port. If not specified, defaults to 8087.  
-H <httpPortNumber> Specify Riak HTTP Port. If not specified, defaults to 8098.  

Concurrency and Misc Settings
--riakworkercount Specify the number of worker threads used to read from/write 
  to Riak (defaults to: 2 * #processor_cores)
--maxriakconnections Specify the max number of connections to maintain
  in the Riak connection pool (defaults to: 2 * workercount, see above)
-v Verbose output, shows number of ops/sec every second. Redundant, default.
-s Turn off verbose output. Only final summary will be output to stdout.
-q Specify the queue size, especially if working with larger object sizes. There are
    at most 2 queues for Load/Dump operations.
    
Copy Settings
--copy Set to Copy buckets from one cluster to another. Cannot be used with d, k, l or delete.
--copyhost <hostName> Specify destination Riak host for *copy* operation
--copyhostsfile <clusterNameFile.txt> Specify a file containing Cluster Host Names.  Req'd
    if a single copyhost not specified.
--copypbport <pbPortNumber> Specify destination protocol buffers port, defaults to 8087.

--destinationbucket <destinationBucket> Specify the destination bucket name for a load or a copy operation.
  Can only be used when specifying a single bucket.

Examples:

Dump (the contents of) all buckets from Riak:

java -jar riak-data-migrator-0.2.9.jar -d -r /var/riak_export -a -h 127.0.0.1 -p 8087 -H 8098

Dump a subset of keys from Riak:

java -jar riak-data-migrator-0.2.9.jar -d -r /var/riak_export --loadkeys bucketKeyNameFile.txt -h 127.0.0.1 -p 8087 -H 8098

Load all buckets previously dumped back into Riak:

java -jar riak-data-migrator-0.2.9.jar -l -r /var/riak-export -a -h 127.0.0.1 -p 8087 -H 8098

Dump (the contents of) buckets listed in a line delimited file from a Riak cluster:

java -jar riak-data-migrator-0.2.9.jar -d -f /home/riakadmin/buckets_to_export.txt -r \
/var/riak-export -c /home/riakadmin/riak_hosts.txt -p 8087 -H 8098

Dump (then contents of) buckets based on the bucket,keys listed in a file, bucket_keys.csv:

java -jar riak-data-migrator-0.2.9.jar -d --loadkeys bucket_keys.txt -r \
/var/riak-export -c /home/riakadmin/riak_hosts.txt -p 8087 -H 8098

Dump (then contents of) a single bucket, A_BUCKET, based on the keys listed in a file, keys.txt:

java -jar riak-data-migrator-0.2.9.jar -d -b A_BUCKET --bucketkeys keys.txt -r \
/var/riak-export -c /home/riakadmin/riak_hosts.txt -p 8087 -H 8098

Export only the bucket settings from a bucket named "Flights":

java -jar riak-data-migrator-0.2.9.jar -d -t -r /var/riak-export -b Flights -h 127.0.0.1 -p 8087 -H 8098

Load bucket settings for a bucket named "Flights":

java -jar riak-data-migrator-0.2.9.jar -l -t -r /var/riak-export -b Flights -h 127.0.0.1 -p 8087 -H 8098

Copy all buckets from one riak host to another:

java -jar riak-data-migrator-0.2.9.jar -copy -r /var/riak_export -a -h 127.0.0.1 -p 8087 --copyhost 192.168.1.100 --copypbport 8087

Copy a single bucket (A_BUCKET) to bucket (B_BUCKET) on the same host

java -jar riak-data-migrator-0.2.9.jar -copy -r /var/riak_export -b A_BUCKET -h 127.0.0.1 -p 8087 --copyhost 192.168.1.100 --copypbport 8087 --destinationbucket B_BUCKET

Caveats:

  • When backing up
  • This app depends on the key listing operation in the Riak client which is slow on a good day.
  • The Riak memory backend bucket listing operating tends to timeout if any significant amount of data exists. In this case, you have to explicitly specify the buckets you need want to dump using the -f option to specify a line-delimited list of buckets in a file.
  • Pay special attention to your buckets and keys, and whether or not URL encoding is needed. When logging to disk, the data migrator will always URL encode buckets.
  • The flags --disable_decoding and --disable_encoding are used to specify what the migrator does when interacting with Riak.

Version Notes:

0.2.9

  • No longer decodes -b (bucket) args

0.2.8

  • Fixed a encoding / decoding issue introduced in 0.2.7.

0.2.7

  • Added a more comprehensive help menu.
  • Added a version option to display the data migrator version.
  • Added two options to disable URL encoding and decoding if needed.

0.2.6

  • Subset of keys can now be dumped, copied, or deleted.
  • In addition to previous capability to specify a file in bucket,key\n format, you can now just have a file with keys as long as a single bucket is also specified

0.2.5

  • Added option to dump a subset of keys

0.2.4

  • Verbose status output is now default
  • Added option to turn off verbose output
  • Logging of final status

0.2.3

  • Changed internal message passing between threads from Riak Objects to Events for Dump, Load and Copy operations but not Delete.
  • Added the capability to transfer data directly between clusters
  • Added the capability to copy a single bucket into a new bucket for the Load or Copy operations.
  • Changed log level for retry attempts (but not max retries reached) to warn vs error.

0.2.2

  • Changed message passing for Dump partially to Events
  • Added logic to count the number of value not founds (ie 404s) when reading
  • Added summary output for value not founds