Skip to content

Latest commit

 

History

History
164 lines (126 loc) · 5.7 KB

index.md

File metadata and controls

164 lines (126 loc) · 5.7 KB

Getting Ready

This guide makes sure that you're ready to work with Maelstrom. We'll install Maelstrom's prerequisites and run a basic test of a pre-existing demonstration server.

Prerequisites

JDK

To run Maelstrom, you'll need a JDK (Java Development Kit), version 11 or higher. If you don't have a JDK yet, you can download one from OpenJDK, or use your system's package manager: e.g. apt install openjdk-17-jdk.

$ java -version
openjdk 14.0.2 2020-07-14
OpenJDK Runtime Environment (build 14.0.2+12-Debian-2)
OpenJDK 64-Bit Server VM (build 14.0.2+12-Debian-2, mixed mode, sharing)

Graphviz

Maelstrom uses Graphviz to generate plots of consistency anomalies. This isn't required for all workloads, but you'll want it for more sophisticated tests. You can install Graphviz from the project site or via your package manager: e.g. apt install graphviz. Once installed, you should be able to run:

$ dot -V
dot - graphviz version 2.43.0 (0)

Gnuplot

Maelstrom uses Gnuplot to generate plots of throughput and latency. You can install Gnuplot from the official Gnuplot site or via your package manager: e.g. apt install gnuplot. Once installed, you should be able to run:

$ gnuplot --version
gnuplot 5.4 patchlevel 1

Installation

Download the latest tarball (maelstrom.tar.bz2, not the source code!) from GitHub, and untar it anywhere. In that directory, run ./maelstrom <args> to launch Maelstrom.

If you'd like to run the latest development builds instead, then install Leiningen(the Clojure build tool), clone this repository, and use lein run <args> instead of ./maelstrom <args>.

Run a Demo

To check that everything's ready to go, you can run Maelstrom against a demonstration program. The program that we'll run is in demo/ruby/echo.rb. Since this program is written in Ruby, you'll need a Ruby interpreter installed.

$ ruby --version
ruby 2.7.2p137 (2020-10-01 revision 5445e04352) [x86_64-linux-gnu]

Let's make sure we can run one instance of the echo server. In the top-level maelstrom/ directory, run:

$ demo/ruby/echo.rb
Online

The program should pause, awaiting input. If you like, you can type or paste a JSON message into the console, like so:

{"src": "c1", "dest": "n1", "body": {"msg_id": 1, "type": "echo", "echo": "hello there"}}

This should print

{"src": "c1", "dest": "n1", "body": {"msg_id": 1, "type": "echo", "echo": "hello there"}}
Received {:src=>"c1", :dest=>"n1", :body=>{:msg_id=>1, :type=>"echo", :echo=>"hello there"}}
Echoing {:msg_id=>1, :type=>"echo", :echo=>"hello there"}
{"src":null,"dest":"c1","body":{"type":"echo_ok","echo":"hello there","msg_id":1,"in_reply_to":1}}

The Received and Echoing lines are informational logging, sent by the server to stderr. The {"dest": ...} JSON message is the server's response to our request, printed on stdout. This is how Maelstrom will interact with our server.

Type [Control] + [c] to exit the server.

Now, let's run Maelstrom, and ask it to test the server:

./maelstrom test -w echo --bin demo/ruby/echo.rb --time-limit 5

This starts a Maelstrom test with the echo workload. Maelstrom will generate requests just like the JSON message we pasted into the server, and read responses back from each server. The --bin argument tells it which program to run for the server. --time-limit 5 means that we'd like to test for only five seconds. This should print something like:

INFO [2021-02-07 16:03:31,972] jepsen test runner - jepsen.core {:perf {:latency-graph {:valid? true},
        :rate-graph {:valid? true},
        :valid? true},
 :exceptions {:valid? true},
 :stats {:valid? true,
         :count 5,
         :ok-count 5,
         :fail-count 0,
         :info-count 0,
         :by-f {:echo {:valid? true,
                       :count 5,
                       :ok-count 5,
                       :fail-count 0,
                       :info-count 0}}},
 :workload {:valid? true, :errors ()},
 :valid? true}


Everything looks good! ヽ(‘ー`)ノ

Maelstrom starts five copies of echo.rb, connects them via its simulated network, and constructs five clients to perform echo requests against those servers. Finally, it checks to make sure that for every request, that server sent back a corresponding echo response. The :stats part of these results tells us that Maelstrom attempted 5 total echo operations, and received 5 echo responses.

This output, along with a bunch of helpful plots and log files, are written to store/<test-name>/<timestamp>/. store/latest is a symlink to the most recently completed test. In that directory, you'll find:

  • node-logs/n1.log, n2.log, etc: The STDERR logs from each node
  • jepsen.log: The full log from the Maelstrom run
  • history.txt: The logical history of the system, as seen by clients
  • results.edn: Analysis results
  • latency-raw.png: Raw operation latencies during the test. Grey regions indicate network partitions.

You can browse these files using a file explorer, at the console, or via a web server that comes built-in to Maelstrom. In the main maelstrom directory, run:

./maelstrom serve
22:41:00.605 [main] INFO  jepsen.web - Web server running.
22:41:00.608 [main] INFO  jepsen.cli - Listening on http://0.0.0.0:8080/

... and open http://localhost:8080 in a web browser to see each test's results.

Congratulations! You're ready to start writing your own systems!