Run like Hell: Docker-Swarm: Running a minimal webserver in a swarm

In my last posting to docker swarm i created a swarm on virtualbox with alpine linux with a hdd footprint of 290MB per node:

There are some tutorials out there with running a nginx or a java webserver in a container but >100MB for each node seems far to much for my tests.

So i decided to create a application which listens on port 8080 with netcat. I created a directory ncweb with ncweb.sh:

ncweb# cat ncweb.sh
#!/bin/bash
sed -i 's/Hostname:.*/Hostname: '$HOSTNAME'/g' index.html
while true; do { echo -e 'HTTP/1.1 200 OK\r\n'; cat index.html;} | nc -l -p 8080; done 2&>1 >> logfile

and a index.html:

ncweb# cat index.html
<html>
<head>
    <title>"Hello, World"</title>
</head>
<body bgcolor=white>
    <table border="0" cellpadding="10">
      <tr>
        <td>
          <h1>"Hello, World"</h1>
        </td>
      </tr>
    </table>
</body>
</html>

The Dockerfile looks like this:

# cat Dockerfile
FROM alpine
WORKDIR /tmp
ADD . /tmp
ENTRYPOINT [ "/tmp/ncweb.sh" ]

After that i created the container:

ncweb# docker build -t ncweb:0.2 .
Sending build context to Docker daemon 4.096kB
Step 1/5 : FROM alpine
---> 053cde6e8953
Step 2/5 : WORKDIR /tmp
---> Using cache
---> c3e11ac3773b
Step 3/5 : RUN mkdir ncweb
---> Using cache
---> d9e634c03cd1
Step 4/5 : ADD . /tmp
---> 95f022aacc1c
Step 5/5 : ENTRYPOINT [ "/tmp/ncweb.sh" ]
---> Running in c1a9e8cee248
---> 6880521f68e4
Removing intermediate container c1a9e8cee248
Successfully built 6880521f68e4
Successfully tagged ncweb:0.2

And let's do a test without docker swarm:

ncweb# docker run -p 8080:8080 ncweb

But running this as a service fails:

# docker service create --replicas=1 --name myweb ncweb:0.2
image ncweb:0.2 could not be accessed on a registry to record
its digest. Each node will access ncweb:0.2 independently,
possibly leading to different nodes running different
versions of the image.
n0himwum38bqzd8ob1vf8zhip
overall progress: 0 out of 1 tasks
1/1: No such image: ncweb:0.2
^COperation continuing in background.
Use `docker service ps n0himwum38bqzd8ob1vf8zhip` to check progress.

and:

# docker service ps n0himwum38bqzd8ob1vf8zhip
ID                  NAME                IMAGE               NODE                DESIRED STATE       CURRENT STATE                 ERROR                        PORTS
8tjsuae9jv8o        myweb.1             ncweb:0.2           node01              Ready               Rejected 3 seconds ago        "No such image: ncweb:0.2"
qp24ssxb5bl5         \_ myweb.1         ncweb:0.2           alpine              Shutdown            Failed 36 seconds ago         "task: non-zero exit (2)"
zwfgcatk7zyi         \_ myweb.1         ncweb:0.2           node01              Shutdown            Rejected about a minute ago   "No such image: ncweb:0.2"
v4a7zkb85yd4         \_ myweb.1         ncweb:0.2           node01              Shutdown            Rejected about a minute ago   "No such image: ncweb:0.2"
ycjftjusv484         \_ myweb.1         ncweb:0.2           node01              Shutdown            Rejected about a minute ago   "No such image: ncweb:0.2"

# docker service rm n0himwum38bqzd8ob1vf8zhip
n0himwum38bqzd8ob1vf8zhip

The error "No such image..." is happening, because the container ncweb is only in the repository of my master.
The easiest way for my test environment is to distribute the local image to all nodes:

# docker save ncweb:0.3 | ssh 192.168.178.47 docker load
The authenticity of host '192.168.178.47 (192.168.178.47)' can't be established.
ECDSA key fingerprint is SHA256:2/8O/SE1fGJ4f5bAQls5txrKMbqZfMmiZ+Tha/WFKxA.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added '192.168.178.47' (ECDSA) to the list of known hosts.
root@192.168.178.47's password:
Loaded image: ncweb:0.3

(i have to distribute a ssh-key to all nodes)

and then:

alpine:~/ncweb# docker service create --replicas=1 --name myweb ncweb:0.3
image ncweb:0.3 could not be accessed on a registry to record
its digest. Each node will access ncweb:0.3 independently,
possibly leading to different nodes running different
versions of the image.

# docker service ps myweb
ID NAME IMAGE NODE DESIRED STATE CURRENT STATE ERROR PORTS
in97xlc7azcw myweb.1 ncweb:0.3 node01 Running Running 8 seconds ago

So my nc-webserver runs on node01, but i can not access it there because i did not define any port mappings ;-(

But finally this command did the job:

# docker service create --replicas=1 --name myweb --publish 8080:8080 ncweb:0.3
image ncweb:0.3 could not be accessed on a registry to record
its digest. Each node will access ncweb:0.3 independently,
possibly leading to different nodes running different
versions of the image.
runf8u9r8719sk13mkf8hh8ec
overall progress: 1 out of 1 tasks
1/1: running
verify: Service converged

The hostname corresponds to the docker container id on node01:

node01:~# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
6c9434b08082 ncweb:0.3 "/tmp/ncweb.sh" 37 minutes ago Up 37 minutes myweb.1.lqiyb34cuxxme2141ahsg8neu

Remaining open points:

Is it possible to do a failback or limit the number of a service per node?
How to get a loadbalancing mechanism for a server application?
(load balancer needed?)
What happens, if the manager fails / is shutdown?

AnonymousApr 16, 2019, 8:48:00 AM
Good information. Lucky me I ran across your website by accident (stumbleupon).
I have saved it for later!
AnonymousApr 3, 2020, 2:15:00 PM
Valuable post useful for everyone.Keep on sharing.
Docker and Kubernetes Training
Docker Online Training
Kubernetes Online Training

Run like Hell

Dec 10, 2017

Docker-Swarm: Running a minimal webserver in a swarm

2 comments: