| tag | ca3d15d5b3d6254a721157146cac3caf88b0fd46 | |
|---|---|---|
| tagger | Tsz Wo Nicholas Sze <szetszwo@apache.org> | Mon Apr 01 23:39:42 2019 +0800 |
| object | 0ca9879f438ffda752a3311a342e85c2ad1cac44 |
ratis-0.3.0 rc0
| commit | 0ca9879f438ffda752a3311a342e85c2ad1cac44 | [log] [tgz] |
|---|---|---|
| author | Tsz Wo Nicholas Sze <szetszwo@apache.org> | Mon Apr 01 20:40:45 2019 +0800 |
| committer | Tsz Wo Nicholas Sze <szetszwo@apache.org> | Mon Apr 01 20:40:45 2019 +0800 |
| tree | f6071f5428a6f1585cb48d918dcc36199547ca8a | |
| parent | bd08c8e7d0d1129f49f04cf3dfbd74490c18c0c2 [diff] |
$mvn versions:set -DnewVersion=0.3.0
Apache Ratis is a java library that implements the RAFT protocol [1]. The Raft paper can be accessed at (https://raft.github.io/raft.pdf). The paper introduces Raft and states its motivations in following words:
Raft is a consensus algorithm for managing a replicated log. It produces a result equivalent to (multi-)Paxos, and it is as efficient as Paxos, but its structure is different from Paxos; this makes Raft more understandable than Paxos and also provides a better foundation for building practical systems.
Ratis aims to make raft available as a java library that can be used by any system that needs to use a replicated log. It provides pluggability for state machine implementations to manage replicated states. It also provides pluggability for Raft log, and rpc implementations to make it easy for integration with other projects. Another important goal is to support high throughput data ingest so that it can be used for more general data replication use cases.
Compile the repository using mvn clean package -DskipTests
To spawn the FileStoreStateMachineServer, use bin/server.sh filestore server --id <selfid> --storage <storage_dir> --peers <id:ip_address,...>
selfid is the id of the instance being spwaned, this should be one of the ids in the peer list.
For example ratis-examples/src/main/bin/server.sh filestore server --id n0 --storage /tmp/data --peers n0:172.26.32.224:6000,n1:172.26.32.225:6001,n2:172.26.32.226:6002
To spawn the FileStoreStateMachine client, use bin/client.sh filestore loadgen --size <file_size> --numFiles <num_files> --peers <id:ip_address,...>
Where, file_size is the size of the file to be generated in bytes, num_files is the number of files to be generated.
For example ratis-examples/src/main/bin/client.sh filestore loadgen --size 1048576 --numFiles 1000 --peers n0:172.26.32.224:6000,n1:172.26.32.225:6001,n2:172.26.32.226:6002
To spawn the ArithmeticStateMachineServer, use bin/server.sh arithmetic server --id <selfid> --storage <storage_dir> --peers <id:ip_address,...>
selfid is the id of the instance being spwaned, this should be one of the ids in the peer list.
For example ratis-examples/src/main/bin/server.sh arithmetic server --id n0 --storage /tmp/data --peers n0:172.26.32.224:6000,n1:172.26.32.225:6001,n2:172.26.32.226:6002
To spawn the ArithmeticStateMachine client, use bin/client.sh arithmetic get --name b --peers <id:ip_address,...>
Where, b is the name of the variable.
For example ratis-examples/src/main/bin/client.sh arithmetic get --name b --peers n0:172.26.32.224:6000,n1:172.26.32.225:6001,n2:172.26.32.226:6002
PS: the peer is a id, ipaddress pair seperated by ‘:’, for eg. n0:172.26.32.224:6000
One can see the interactions of a three server Ratis cluster with a load-generator running against it by using the run_all_tests.sh script found in [dev-support/vagrant/]. See the [dev-support/vagrant/README.md] for more on dependencies and what is setup. This will allow one to try a fully setup three server Ratis cluster on a single VM image, preventing resource contention with your development host and allowing failure injection too.
[1] Diego Ongaro and John Ousterhout. 2014. In search of an understandable consensus algorithm. In Proceedings of the 2014 USENIX conference on USENIX Annual Technical Conference (USENIX ATC'14), Garth Gibson and Nickolai Zeldovich (Eds.). USENIX Association, Berkeley, CA, USA, 305-320.