Last modified: 2014-03-08 18:35:24 UTC
Puppet changes are causing issues on labs for a number of reasons - a good example of this is the ircecho module. -- Things that move/rename/change existing classes break labs instances * Wikitech is not synced with git - this means classes moving such as misc::ircecho to ircecho do not get updated in the database or ldap. - Breaking wikitech is bad, but workable. - Breaking puppet runs on instances is really bad. - Migrations for the db (for wikitech) and ldap (for the enc) should be performed after git changes - This should be an automated process * Using maps is not possible on labs - Limitation in the ENC AFAIK - Simple migration of classes is not possible -- Changes should be tested and deployed in labs before production * Numerous bugs have appeared in labs due to prod changes and prod after ignoring labs issues - Workflow should be more enforced -- Test in labs -- Deploy to labs -- Deploy to prod * Issues in relation to labs and prod running the same branches - Can't test major changes using puppetmaster::self sensibly - De-sync between production and labs while sharing a git repo causes issues - Testing on labs ensures changes don't break labs or production * Automated testing should be possible - Lack of API on wikitech for creating instances - Jenkins should be able to create set instances and run tests -- Beta on demand (different branches etc) -- Changes such as the ircecho one would be clearly picked up etc Currently the whole workflow/experience thing is just horrible imo.