Deployment tooling/Cabal/2015-07-27
2015-07-27
edit- Marko playing with Scap 3-point-oh
- this looks like some great stuff, Tyler!
- does it do a rolling deploy or parallel?
- same question for service restarts
- Tyler: currently runs parallel, but should be simple to serialize/make that an option
- for storage changes: we need to start only one process on one node to apply it, destroy it and then do a normal restart
- problem: worker restarts
- context: the master RESTBase process monitors its workers, when too much RAM, the worker is killed and respawned
- this implies that some workers might start using the new code before an actual service restart
- this is a problem for storage schema changes perhaps, need to discuss
- potential problem: git pull needs to be done before issuing the 'deploy' command
- perhaps a param in scap.cfg or command line could tell scap which sha1 (or branch name) to check out and do that before tagging and sending the code?
- Ansible initial thoughts/goals
- Wrapping up a standard deploy via ansible seems fairly simple (Ansible is built with ansible...so...)
- Started on working with it
- End goal
- A standard deploy, rollback, etc without having to learn any ansible-domain-specific code
- Allow to extend (post-fetch, post-deploy actions, etc) via ansible-domain-specific code
Last week's goal
edit- Play with Scap 3.0/Trebuchet
- Rank them on the Spreadsheet
- Tyler/Mukunda: Work on scap improvments (we didn't)
- Tyler: Work on Ansible (did some)
Topics
editWeekly goal
editEtherpad backed up to https://www.mediawiki.org/wiki/Deployment_tooling/Cabal Spreadsheet: https://docs.google.com/spreadsheets/d/1MlEsFxrLvdZdV_G82WEAIvBXr7ArO7nCEKaFClHhJEw/edit#gid=0