- old orchestration
pick vlan, compute, storage
vms (linked clone + puppet/chef/etc for infra specific changes) + necessary packages for the app
hand over to the app admin teams to deploy the app (deploy)
app vs service distinction
- new orchestration
deployment: pull the code from git (heroku, openshift, flynn, dokku)
: pull the code as tar.gz (called blob store in cf)
: compile the code locally (worker vms in cf) and copy the exec to the Paas
provisioning: (done by bosh)
set up containers (DEA, agent)
config networking
install/config OS services or libraries like apache, ruby gem (buildpacks in cf)
lifecycle mgmt:
start the app
monitor/repot
restart
stop/start cli
service: polyglot persistence
service broker (managed): register/add/delete mysql/oracle/redis/rabitmq/etc
user-managed (legacy)
CF ops manager (GUI); behind the screens (bosh)
Bosh: stemcell + release + config.yml/manifest
bosh cli (esxcli) --> bosh director (vcenter) --NATS--> tasks
manifest: cloud infrastructure + n/w infra + vm types + OS type
Stemcell + Release + Manifest (SRM)
Changing clouds:
change the template/stemcell + change manifest (SM)
scale up app:
change one line in the manifest (M)
update/rollback the app:
change the release (R)
No comments:
Post a Comment