DevOps 2018
The Past and The Future
Actually
We had a plan
- Blue/Green deployments
- Monitoring on steroids
- Support for A/B testing and Feature Toggles
This happened
Problems
- Lack of knowledge
- Lack of human resources
- Maintenance of existing stuff
- Lack of focus
When you switch context too often
Achievements
- DevOps hire
- Fixed dozens of issues (e.g. check monterail/ops and monterail/ansible-XYZ repos)
- Learned A LOT
- Real plan start emerging
Plans
Automate all the things!
- infrastructure-as-code
- configuration management
- service discovery
Automate all the things!
infrastructure-as-code
- Terraform
- infrastructure described in executable manifests
- infrastructure code stored along side application code
- environment reproducibility
Automate all the things!
configuration management
- Ansible
- already much done
- server configuration stored along side application code
- avoid "configuration drift"
- environment reproducibility
Automate all the things!
service discovery
- Consul
- scale-ready
- no hard-coded service endpoints
- health checks
Automate all the things!
infra-as-code + config mgmt + service discovery
- Kubernetes
- multi-node Docker cluster
- declarative manifests
- built-in service discovery
- entire datacentre as one single compute unit
- simplified monitoring
Standardise!
- policies and SOPs
- "wtf is sop?" -- Standard Operating Procedure
- DevOps <=> ITIL
- post mortems
- transparency and accountability
Questions?
Thank you!
DevOps 2018
By Jan Dudulski
