pipelines
Thursday, Stage, 13:00
- richard lavey / Facilitator
- Scott Russell
Thoughtworks go, used, but it is always good to sit down and write/draw out your pipeline. 5 year trip. Bad, but now bad, but better than it was, but not right.
How do you want the human being's to behave. eg, any developer can ship at any time and be safe. How do you design a system to do this. What are the rules for who fixes it.
When drawing out the pipeline,
The fast and safe part is difficult, fast is easy. Fast and stable, but building slow and stable on the same system is Haard!.
All the safety systems rely on the system acting at a specific speed.
Try to think about one vertical slice of the org. Business to infrastructure. Try not to call the things "production", so that no-one owns it. This is the strategic viewpoint, then move to tactical. Don't compromise the architecture and sacrifice the speed.
It is a long time, and start with a slice of the organistaion. First, CD is a vertical slice to transform.
Buyin from senior management requires buyin from senior managers, and celebrate success early.
Why do you go faster, because it is fun, so need a reason, or a business case.
Failure to articulate the reason is a common failure mechanism.
A dishonest person in the pipeline will also break the pipeline.
Failure to understand that the system you are changing is to change human behaviour. - easy to forget.
If the system reinforces the behavior changes, then it has a higher chance of success.
Individual's take responsibility for the outcomes.
Have to empower the individuals, this is difficult, but it is where you have to get to.
Q: Where does Chef fit in this pipeline A: Chef is good as the execution framework. Tomorrow, it has to be good at the workflows. half a dozen at the most Design one pipeline well and use this to bring the company forward? There is the system that is in place, but you measure it by the outcomes of their behavior. There is magic in the workflow, but not in the tooling. The workflow is not arbitrary, and it is not really flexible.
Q; Propogation of things through environments, how to? A: Artifactory, and jenkins given as examples. Using tags or buckets. Drive the workflows with something other than Chef. Important to understand that the screen view of the pipeline, is not what matters, it is the understanding Coupling of infrastructure and application is recommended, as at speed they are tightly coupled ( but this does not mean all your infrastructre is at the same patch level, but it might mean for a given application that the infrastructure is at the same level ). Application and infrastructure Allow people the freedom to change everything, but concentrate on the failure changes.
Use these names in stead of dev/test/prod etc. Verification, release, delivery, etc
Organisational startegy, tell me all the ways things fail, and all you have to do is fix all the things,job done.