A few days ago, I was officially introduced to Couchbase at a Toronto Hadoop User Group meetup. I say “officially” because I’ve known about some Couchbase use-cases / pros and cons on a high level since v1.8, but never really had the time to look at it in detail.

As the presentation progressed, it got me interested in actually tinkering around with a Couchbase cluster (kudos to Don Pinto). My first instinct was to head over to the Docker Registry and do a quick search for couchbase. Using the dustin/couchbase image, I was able to get a 5-node cluster running in under 5 minutes.

Run 5 containers

1
2
3
4
5

docker run -d -p 11210:11210 -p 11211:11211 -p 8091:8091 -p 8092:8092 --name cb1 dustin/couchbase
docker run -d --name cb2 dustin/couchbase
docker run -d --name cb3 dustin/couchbase
docker run -d --name cb4 dustin/couchbase
docker run -d --name cb5 dustin/couchbase

Find Container IPs

Once the containers were up, I used docker inspect to find their internal IPs (usually in the 172.17.x.x range). For example, docker inspect cb1 returns

1
2
3
4
5
6
7
8
9
10
11
12

[{
    "Args": [
        "run"
    ],
    "Config": {
        "AttachStderr": false,
    ...
    "NetworkSettings": {
        "Bridge": "docker0",
        "Gateway": "172.17.42.1",
        "IPAddress": "172.17.0.27",
    ...

Update: Nathan LeClaire from Docker was kind enough to write up a quick script that combines these two steps:

1
2
3
4
5
6
7
8
9

docker run -d -p 11210:11210 -p 11211:11211 -p 8091:8091 -p 8092:8092 --name cb1 dustin/couchbase

for name in cb{2..5}; do 
    docker run -d --name $name dustin/couchbase
done

for name in cb{1..5}; do
    docker inspect -f '{{ .NetworkSettings.IPAddress }}' $name
done

Setup Cluster using WebUI

If cb1 is at 172.17.0.27, then the Couchbase management interface comes up at http://172.17.0.27:8091 and the default credentials are:

1
2

Username: Administrator
Password: password

Once you’re in, setting up a cluster is as easy as clicking “Add Server” and giving it the IPs of the other containers. As soon as you add a new server to the cluster, Couchbase will prompt you to run a “Cluster Rebalance” operation – hold off until you’ve added all 5 nodes and then run the rebalance.

Push some data into the cluster

Once the cluster was up, I wanted to get a feel for how the WebUI works so I wrote this script to grab some data from our existing cluster of JSON-store-that-I-am-too-ashamed-to-mention and added it to Couchbase:

1
2
3
4
5
6
7
8
9
10
11
12
13

#!/usr/bin/env python
# encoding: utf-8

from pymongo import MongoClient
from couchbase import Couchbase

src = MongoClient(['m1', 'm2', 'm3'])['twitter_raw']['starbucks']
tar = Couchbase.connect(bucket='twitter_starbucks',  host='localhost')

for doc in src.find():
  key = str(doc['_id'])
  del doc['_id']
  result = tar.set(key, doc)

The Couchbase python client depends on libcouchbase. Once those two were installed, and the twitter_starbucks bucket had been created in Couchbase, I was able to load ~100k JSON documents in a matter of minutes.