Docker Swarm Implementation

[diemol]

Comes from #907, thanks to @yosserO and @tstern
Running all CI

[codecov-io]

Codecov Report

Merging #982 into master will decrease coverage by 4.01%.
The diff coverage is 3.95%.

@@             Coverage Diff             @@
##             master    #982      +/-   ##
===========================================
- Coverage     58.51%   54.5%   -4.02%     
- Complexity      642     646       +4     
===========================================
  Files            54      56       +2     
  Lines          4127    4455     +328     
  Branches        370     413      +43     
===========================================
+ Hits           2415    2428      +13     
- Misses         1493    1804     +311     
- Partials        219     223       +4

[diemol]

@yosserO @tstern
Things look good to me, I am just missing documentation for end users. What we can do is to merge this to master, release the feature as beta, and in a subsequent PR you could add documentation.

I'll also appreciatte if you can hangout on the #zalenium channel in the Selenium Slack, in case people have questions.

What do you think?

[yosserO]

Hi Diego,
Thank you for reviewing the PR.
We will create a PR for the documentation next week. We are already in the zalenium and Selenium Slack Channels. If there are questions related to docker swarm, please just mention us in the comment so we can get a notification.

[trinhpham]

Really eager to be one of the first users :)
My question is: when and where can I get the beta build?

[trinhpham]

The approach of creating new service when need to create new container is not as good as pre-defining the service then update it scaling.
Refer: https://docs.docker.com/engine/swarm/swarm-tutorial/scale-service/

[trinhpham]

I think it should be better if we pre-define the worker services, then update theirs scaling like https://docs.docker.com/engine/swarm/swarm-tutorial/scale-service/

[yosserO]

Hi @trinhpham ,

Service scaling was our first Approach for creating and deleting Browser tasks.

The problem was: When scaling down the number of replicas, it doesn't took in account which task is not active. It killed containers that were actually running tests and made the tests fail. We found no way to specify the task to shutdown explicitly when scaling down the number of replicas.

moby/swarmkit#2264

If we scale the number of replicas up instead of creating new services. We will have to scale them down when the tests end. And this will lead to a wrong behaviour.

[trinhpham]

Got your idea. Thanks for explaining it to me :)

[diemol]

Ok, let's merge this and do a release!

[yosserO]

@trinhpham

Really eager to be one of the first users :)
My question is: when and where can I get the beta build?

We created a pull request to add documentation: #984