Skip to main content

Playing with ETCD cluster in Docker on Local

I've started to write some management component last week. We would like to utilize CoreOs with the whole stack, as much as possible, at least within such early phase of our project.

The core component of our solution is ETCD - distributed key value store. Something like my favorite piece of software - Redis. Word 'distributed' means that the core of all things within your solution needs to be synchronized or 'consensused'. ETCD uses Raft. I'd love to know how my desired component works in real environment where everything can die.

In the age of docker - where every piece of software is docker-ized, it's pretty simple to start ETCD cluster on local in a second. Following piece of code starts three etcd instances linked together in one cluster.

docker run -d -p 4001:4001 -p 2380:2380 -p 2379:2379 --net=host --name etcd0 \
 -name etcd0 \
 -advertise-client-urls http://localhost:2379,http://localhost:4001 \
 -listen-client-urls http://localhost:2379,http://localhost:4001 \
 -initial-advertise-peer-urls http://localhost:2380 \
 -listen-peer-urls http://localhost:2380 \
 -initial-cluster-token etcd-cluster-1 \
 -initial-cluster etcd0=http://localhost:2380,etcd1=http://localhost:2480,etcd2=http://localhost:2580

docker run -d -p 4101:4101 -p 2480:2480 -p 2479:2479 --net=host --name etcd1 \
 -name etcd1 \
 -advertise-client-urls http://localhost:2479,http://localhost:4101 \
 -listen-client-urls http://localhost:2479,http://localhost:4101 \
 -initial-advertise-peer-urls http://localhost:2480 \
 -listen-peer-urls http://localhost:2480 \
 -initial-cluster-token etcd-cluster-1 \
 -initial-cluster etcd0=http://localhost:2380,etcd1=http://localhost:2480,etcd2=http://localhost:2580

docker run -d -p 4201:4201 -p 2580:2580 -p 2579:2579 --net=host --name etcd2 \
 -name etcd2 \
 -advertise-client-urls http://localhost:2579,http://localhost:4201 \
 -listen-client-urls http://localhost:2579,http://localhost:4201 \
 -initial-advertise-peer-urls http://localhost:2580 \
 -listen-peer-urls http://localhost:2580 \
 -initial-cluster-token etcd-cluster-1 \
 -initial-cluster etcd0=http://localhost:2380,etcd1=http://localhost:2480,etcd2=http://localhost:2580

The inspiration is obvious, but this stuff simply runs everything on your computer.  Parameter --net=host provides full transparency from port&network point of view.

You can now use following URL in a browser:


Good thing is also to check all members of your cluster. You will kill them later.


You can easily delete all keys in XYZ namespace using curl once you did you tests. Note that you can delete only one of your keys so you can't perform following command on your root namespace.


I also prefer to see http status code as ETCD uses http status codes.

curl -v

In advance to status codes, it always returns a json with their own errors codes. See a snippet at the end of the following listing. You can get something similar to:

* Hostname was NOT found in DNS cache
*   Trying
* Connected to localhost ( port 2379 (#0)
> GET /v2/keys/XYZ HTTP/1.1
> User-Agent: curl/7.35.0
> Host: localhost:2379
> Accept: */*

< HTTP/1.1 404 Not Found
< Content-Type: application/json
< X-Etcd-Cluster-Id: 65a1e86cb62588c5
< X-Etcd-Index: 6
< Date: Sun, 01 Mar 2015 22:55:14 GMT
< Content-Length: 69

{"errorCode":100,"message":"Key not found","cause":"/XYZ","index":6}
* Connection #0 to host localhost left intact

At the end of playing with ETCD cluster, you will probably want to remove all etcd's containers. I use simple script which removes every docker container, but you can improve it using grep to remove only those hosting ETCD.

sudo docker rm -f `docker ps --no-trunc -aq`

The last interesting thing is the performance. I've reminded Redis which can handle one million of transactions per second using one thread. I was surprised when ETCD responded usually in 20-30ms. Much worse fact is that I've also encountered client timeouts because of 400-500ms RT per request. Raft is obviously not for free. But the purpose of ETCD is massive reading scalability. Well, good to know.

Popular posts from this blog

Http and TCP Load Balancing with Kubernetes

Kubernetes allows to manage large clusters which are composed of docker containers. And where is large computation power there is large amount of data throughput. However, there is still a need to spread the data throughput so it can reach and utilize particular docker container(s). This is called load balancingKubernetes supports two basic forms of load balancing. Internal among docker containers themselves and external for clients which call you application from outside environment - just users.

Internal Load Balancing with Kubernetes Usual approach during the modeling of an application in kubernetes is to provide domain models for pods, replications controllers and services. Look at the great documentation if you are not aware of all these principles.

For simplification, pod is a set of docker containers which are always located on the same node. Replication controller allows and guarantees scalability. The last but not least is the service which hides all instances of pods - cre…

Validating nginx config file using docker approach

I try to setup nginx as a load balancer. The configuration is just a file with a lot of constrains so I need a validation. There is no online validation service, as e.g. CoreOS has, and I don't want to install nginx on my laptop as I work on a distributed app.

Docker is right approach for me. Let say I have following config:

In short, I'm going to pass nginx config to running nginx instance and look to the logs.

Put you nginx.config to the temp and start the docker image:

sudo docker run --name nginx -v /tmp/nginx.config:/etc/nginx/nginx.conf:ro -d nginx It uses volume mapping so the command just starts a new docker container and mounts a local /tmp/nginx.config to the given in-container path. You can obviously change the volume path to your personal path. Is it working or not? Look at logs.

sudo docker logs nginx If there is no entry, your file is fine. In the case of an error, you can see something like this:

2016/01/08 11:37:31 [emerg] 1#1: unexpected "}" in /etc/…