The press is mixed on Docker (and to a lesser extent rkt), especially for use in production environments. Is anyone finding success with Docker or rkt as part of their production deployment?
I work for an extremely large - £Billions - retail grocery operation in the U.K. We use Docker containers in production, not right across the business, but in many important and up-and-coming areas. What did you want to know?
I was really looking to take an informal poll to see how common containerization is in production, but now that you've asked, I guess my questions revolve around the technology stack:
Are you using kubernetes for orchestration?
Is there automation managing the orchestration configurations like Puppet, Chef, Ansible, Saltstack?
Can I assume this is being carried out through virtualization, if so, what kind of hypervisor is in use?
Is this on premise or in the cloud?
More questions to come, I'm sure and thanks in advance for indulging my curiosity!
Yes on Kubernetes - it's not a platform in and of itself, but an excellent place to found a platform.
Automation everywhere. Chef and Ansible, but not on the same project.
It's all AWS-based. We do use ECS on occasion but it's not great.
As others have said, the main issues are developing the entire CI/CD pipeline.
We're running significant loads on Azure with Mesos+Docker (Azure Container Service).
Mesosphere's stack removed the Docker runtime and only uses Docker for image packaging. Stable and in production for over 8 months now.
Haven't used AWS/Google yet for the same set up, we've customized the Azure set up to include tens of TBs of disk space, over 100 compute nodes with quite a bit of RAM. I'd say for any reasonably sized start up or company, it's a good choice - plus the Mesosphere stack also helps avoiding lock-in to one vendor. (would love to see them making multi-cloud tooling)
I don't plan to at my own company, and I certainly wouldn't have done it anywhere until recently, but am currently setting up docker for CI/CD and production for a side client whose previous development was a total trainwreck India outsource jobbie with zero oversight and four years of extreme technical debt.
I am not sold on docker per-se, but it is a reasonably easy choice for a container system available to demand foreign developers on arbitrary platforms standardize on testing with. Containers are auto-generated by the CI/CD system, based on git commits. We're also using gitlab-ce for this.
I work for a well known software company on a SaaS product. We've been using Docker in production for around 2 and a half years, and started integrating with docker in earnest close to 3 years ago. We were mostly drawn to the ease of creating resource isolated processes. We run in AWS on our own brain dead "orchestration" system but we are looking to use something purpose built soon.
To give a sense of scale we have over 1000 containers running at any given time and go through ~15k per day since our containers don't live for very long.
I'm not sure what press is mixed but the industry as a whole has certainly embraced containers. I'd be suspect of any press that determines containers are vaporware and not worth the time.
14 comments
[ 3.3 ms ] story [ 39.3 ms ] threadAre you using kubernetes for orchestration?
Is there automation managing the orchestration configurations like Puppet, Chef, Ansible, Saltstack?
Can I assume this is being carried out through virtualization, if so, what kind of hypervisor is in use?
Is this on premise or in the cloud?
More questions to come, I'm sure and thanks in advance for indulging my curiosity!
Disclaimer: I deploy dc/os for customers on prem alongside our paid product.
I am not sold on docker per-se, but it is a reasonably easy choice for a container system available to demand foreign developers on arbitrary platforms standardize on testing with. Containers are auto-generated by the CI/CD system, based on git commits. We're also using gitlab-ce for this.
Multiple iterations on it
1. Orchestration through chef, no service discovery, Just containers on machines as micro-services.
2. Orchestrations through mesos and marathon.
3. Kubernetes and Rundeck.
#3 has been the best in terms of flexibility, #1 has been the best in terms of ease of use and fire-and-forget.
The most complicated part about Docker in production is not actually Docker in production it's everything on the way.
From CI to building to publishing and deployment.
To give a sense of scale we have over 1000 containers running at any given time and go through ~15k per day since our containers don't live for very long.
I'm not sure what press is mixed but the industry as a whole has certainly embraced containers. I'd be suspect of any press that determines containers are vaporware and not worth the time.