gistfile1.txt

Jenkins Evolutions
Toomas Römer (dow.ngra.de / @toomasr )
---
Test -> CI -> CD 

Why evolutions:
* authentication
* auth
* logging
* notify
* remote machines

"Cron on stereoids", "unopinionated build tool"

https://github.com/jenkinsci/jenkins-scripts ( job hierarchy )

zero turnaround custom dashboard.. Looks nice.


jenkins monitoring:
* cluster needs to be top prioritty just as other infrastructure
* from baremetal -> cloud -> back to desktop machines (because of IO load)
* cloud: 1 small isntance per JDK, manual update of ami, slow, scales

SSD -> 30% speed increase



servlet tests
---
1) start container (jboss, tomcat, ...)
2) deploy apps
3) test resulst, reload changes, test results


exclusive conainer access: throttle concurrent builds plugin

jboss arquillian + jenkins = ?



DVCS
----
ZT uses hg

problem
* many jdk
* many app servers
* many servlet containers

slow feedback cycle

difficult to install environments

split tests:
* dev branch: small test suite
* if green -> run large test suite



general stuff
----
Fight for executors ( 2 release on same day)

Optimum utilization should be the goal

Quick feedback is grand goal

-------------

* no staging jenkins. 


Questions
---
* tried local cloud? eucalyptus? openstack?
amazon ec2

* fail fast: could you run previously failing tests first?
nope

* fight for executors: solved by cloud overflow?
works fine