Building Trust with Continuous Delivery

By: Chris Weldon

Transcript

1. Continuous Deployment to Build Trust CHRIS WELDON SENIOR CONSULTANT IMPROVING ENTERPRISES
2. Chris Weldon u  Senior Consultant u  Microsoft MVP u  Slings Code in C# and PHP u  Systems Administrator devops u  chris@chrisweldon.net u  @neraath
3. And now for something completely different
4. As a conference presenter, I would like a web site with upcoming conferences, so I can stay in-the-know on the conference circuit.
5. As a conference presenter, I would like to see a list of upcoming conferences, so that I know what to submit presentations for.
6. Our Journey Current Continuous Deployments
7. In Business
8. We Develop Software
9. When developing software…
10. We often try as a team…
11. But are often racing the clock…
12. So we rush Apps to Production.
13. This cycle repeats and we make Frequent Changes to Production.
14. And Our Apps Repeatedly Break
15. When stuff breaks… The business gets mad!
16. They start to see us as “rogues”
17. And institute gates to stop this “bad” activity
18. Gates are typically nothing more than
19. Why?
20. It’s about Trust.
21. How do we rebuild trust?
22. Back to basics.
23. We write software to support the business.
24. Or software is the business.
25. We need to adapt to changes in business needs.
26. We need to (rapidly) fix bugs.
27. We properly test first.
28. In order to maintain stability…
29. …we should have slow, continuous evolution of features.
30. So how do we support these goals?
31. Strive for technical excellence!
32. Remember to test!
33. We should constantly integrate all changes back into mainline.
34. requent Deployments Amount of Change Inf F
35. Automate Deployments
36. Our Journey Current Continuous Deployments Automation
37. Technical Excellence
38. Let’s talk about your databases…
39. As a conference organizer, I would like conferences I create to persist beyond my session so that other users can see upcoming conferences.
40. Is your DBA’s idea of SQL Management pitching him a SQL script to run?
41. That’s not automation.
42. Solution: Migrations u  Entity Framework Code-First Migrations u  Migrator.NET u  RoundhousE u  And many more…
43. Migrations Demo Demo
44. Configuration Management
45. As a systems administrator, I would like enable caching of conferences from the database so that we can minimize the impact of querying the database.
46. Scenario: Multiple web applications u  CritterMatch Core Business Libraries (CMCBL) u  CritterMatch.com depends on: u  log4net v1.0.6 u  Entity Framework v4 u  ASP.Net Web API v1.0.0 u  CMCBL v1.0.0 u  CritterMatch Business Services depends on: u  log4net v1.2.1 u  Entity Framework v5 u  ASP.Net Web API v1.2.2 u  CMCBL v1.1.0
47. NuGet
48. Scenario: Multiple web applications u  CritterMatch.com u  Developer Workstations u  Integration/QA u  Staging u  Production u  CritterMatch Business Services u  Developer Workstations u  Integration/QA u  Production – Client A u  Production – Client B
49. Configuration Management u  Configurations likely to change: u  Database Connection Strings u  Web Service Connection Endpoints u  Authorized Users? u  Requirements Management u  Different views or behaviors u  Dependency Injection? u  Often times a configuration change, too.
50. Configuration Management Strategies
51. Logging and Monitoring Strategies
52. Better Know a Framework
53. Monitor all teh things! u PIC: Dashboard rather than text u CPU, Memory, Disk I/O, Network I/O u Active Users, Requests per Second, Database Queries per Second u # of 200, 300, 400, 500-level requests per second u Heuristics
54. PIC: Alarm
55. PIC: Scaling
56. PIC: False Positive
57. Making Logging Useful to Ops & Engineering u  Ops needs to know: u  Network connection problems u  Load problems u  Service/dependency outages u  Consider surfacing events to event log for ops, leaving log files to devs u  Engineering needs to know: u  Everything that ops needs to know u  Performance
58. Debug is never for production u  Not constantly, that is. u  If you are trying to debug the production database query, you’re doing it wrong. u  If you are trying to figure out the object model being sent to the view, you’re doing it wrong. u  If you are trying to debug the workflow of your application, you’re doing it wrong. u  The highest level of constant production logging should be INFO. u  Info serves for interactions between customers and your site. u  Info surfaces information customers have requested, but never the interactions of your code.
59. Exception Management u  Don’t always surface exceptions to your Event Log if Ops is monitoring u  Boy Who Cried Wolf u  Be meticulous about which exceptions are surfaced out of your application u  Exceptions are expensive
60. Demo
61. Deployment Techniques
62. Continuous Deployment = Release to Production After Every Commit?
63. We don’t want to be rogues…
64. Release Management u  How to manage multiple environments? u  Builds for Each Environment u  Versioning
65. devops u  The collaboration, communication, and integration between developers and operations to reduce the friction of getting software releases into production.
66. Production Push/Pull
67. CI Push
68. But how do we establish trust?
69. Our Journey Current Continuous Deployments Automation Testability
70. Unit Tests?
71. What about your UI?
72. What about dependency or component failure?
74. Smoke Tests!
75. Behavioral/Acceptance Tests
76. Still having trust issues?
77. Our Journey Current Continuous Deployments Automation Testability Manageable
78. Gated deployments
79. Rollback Techniques
80. Let’s see it!
81. Test labs
82. Behavior Driven Infrastructure
83. Our Journey Current Continuous Deployments Automation Testability Manageable
84. Recap
85. Q&A
86. Ratings, Please! http://spkr8.com/neraath