WarStoriesAndSuccesses: Difference between revisions
From CitconWiki
Jump to navigationJump to search
Created page with "Test strategies and flow and sharing of strategies * Test pipelines and integration strategies across teams * Architecture * It’s so important to assume you will miss somet..." |
No edit summary |
||
| Line 8: | Line 8: | ||
* Defeat: why can’t anyone deploy anything | * Defeat: why can’t anyone deploy anything | ||
* Victory: 3 months to change a web page —> changes in 1 day | * Victory: 3 months to change a web page —> changes in 1 day | ||
** Text link to 404 for the purpose of recording clicks of interest for house viewing | |||
* PJ: Company daily release process but often would break and often take 3 people most of the day to get it out | * PJ: Company daily release process but often would break and often take 3 people most of the day to get it out | ||
** Eventually, every time someone made a change of the code it would be deployed to prod and monitored | |||
*** End to end test bankruptcy | |||
** Success | |||
** Monolithic system (with seams but tight coupling) was broken down into component based independently deployable systems | |||
** Be prepared, this process will take 3 years (or more) | |||
** How do you sell this idea? | |||
*** The key to our success is shortening our feedback cycles | |||
*** From Tony later: show other companies that are already past 3y to show what’s possible (and build trust) | |||
* Tim: traditionally siloed functional org | * Tim: traditionally siloed functional org | ||
* Idea to cash was ~ 200d | * Idea to cash was ~ 200d | ||
Revision as of 04:11, 18 May 2019
Test strategies and flow and sharing of strategies
- Test pipelines and integration strategies across teams
- Architecture
- It’s so important to assume you will miss something .. so MTTR is critical
War stories
- Defeat: why can’t anyone deploy anything
- Victory: 3 months to change a web page —> changes in 1 day
- Text link to 404 for the purpose of recording clicks of interest for house viewing
- PJ: Company daily release process but often would break and often take 3 people most of the day to get it out
- Eventually, every time someone made a change of the code it would be deployed to prod and monitored
- End to end test bankruptcy
- Success
- Monolithic system (with seams but tight coupling) was broken down into component based independently deployable systems
- Be prepared, this process will take 3 years (or more)
- How do you sell this idea?
- The key to our success is shortening our feedback cycles
- From Tony later: show other companies that are already past 3y to show what’s possible (and build trust)
- Eventually, every time someone made a change of the code it would be deployed to prod and monitored
- Tim: traditionally siloed functional org
* Idea to cash was ~ 200d * 91% wait time * First response was to be mad at the person pointing at the problem * Improvement is slow, but it is improving * It was important to just get people talking about it even if they got upset initially * Value stream map made it all possible
- Success: Operations team and developer teams
* Devs write run-books over a couple months, then throw it to the operators
* Big bang release every couple of months
* Build a more intimate relationship between Dev and Ops
* Moved to central Jenkins instance from individual snowflakes
* This integrated the dev workflow with the ops workflow (common environment)
* Decoupled infra track from application track for deployments
* Over a 6 month period — still monthly releases but far less stressful on everyone involved
* Enjoyable pizza
- Frederick: Goal: faster release cycle — get to weekly
* Current was monthly with bi-weekly patch releases * Visualized value stream maps with both * Took similarities and combined to 1 flow like the patch flow — currently at every 2w * Giant monolith * There were a lot of fears and resistance but enough trust was built to give it a try * Now trying to work towards TBD
- Small web team gets security scan result
* Pen testers test quarterly — first report was very ugly
* Introduced PR process, but devs weren’t able to spot security vulnerabilities
* Secure Code Warrior gamefies this .. points for vulnerabilities found
* Got some training (e.g. OWASP) and testers were more involved
* Making progress
* “Why should I train you (freelancer), what if you leave"
* “What if you don’t train us and we stay” (response)
- Failure:
* Cultural, hierarchical issues with a Dev silo, QA silo (which moved closer) and DevOps (eww) silo which was moved even further away * Improvements were made in delivering more often, but then there was some realization/questioning of this and so deliveries are now getting slowed/spaced out