author: Dan Chuparkoff
Dan will present how, in less than 6 months and with a 1-person team, they went from no infrastructure automation, to having all of their infrastructure automated with Ansible. Learn how BigPanda handles zero-downtime infrastructure updates and connects ansible with their chat infrastructure, and some strategies on managing automation projects with very small teams.
viewers: 140 Million with an M comercial # 1 airs Instant crash since so many people were trying to get free tickets! Re-run comercial! 100,000 free tickets in 42 seconds authorized to give out 300,000 tickets!
Universal Studio guy said something like:
just get more servers! and run it again!
Create infrasturture around The one man:
No matter what the complexity of the tesk it boils down to running and ansible module, which makes playbooks easy to write and easy to review later.
run as many times as you want and it won't hurt anything!
Every script we write is used 2 times once when deploying a new version and again when provisioning a new server using tags we run our skip tasks or plays as needed:
./playbook.yml -i hosts -- --tags infra
# infra could be replaced with run_once
Can be easily extended with JSON Correlating monitoring alerts with application deployments reduces MTTR MTTR: Mean Time to Repair
No one touches servers, go through ansible
Leverage peer review behavior Bring agile dev stuff to the ops world Made into a meetup bigpanda/ansible-workshop
# People can make mistakes here!
ansible-playbook playbook.yml -i hosts -e var1=val1 -e var2=val2
Make it better! bigpanda/ansible-exec
deploy bigpanda
There are a very small number of commands like this in bigpanda Learn the hard stuff one time and make sure the playbook is perfect Kept healthy through peer reveiw!
Ansible at this point is so easy, everything is an edit from a playbook
bigpanda uses hipchat Deployment talk happens here BeanBot listens to hipchat and waits for something like
Hey Beanbot!
BeanBot - ender's game reference to Ender's friend Bean Based from HuBot
- Able to replace manual deploys
- ensured that everything was configured exactly as intended
- 100% ansible created all with zero downtime
Now that everything is in ansible We get a changelog for our inrfrastructure There is an ansible plugin for bigPanda You get a list of alerts that occur for your ingfrastructure This bundles problems together
Q: You Correlate what was changed and who changed it? A: User id is passed when you do a beanbot deploy thus ownership is established