Open adamsiemion opened 6 years ago
I would find it more useful to change the flag to an enumeration e.g. -old-app=remove|stop|none where remove is the default, none solves your case and stop solves my case https://github.com/contraband/autopilot/issues/53. What do you think?
I'm willing to help, unfortunately new to the language go.
@mibollma Good work on the code changes! I support the idea. Actually recently we have moved away from this plugin to the tool that does the same thing but just calls cf
with appropriate commands (https://github.com/adamsiemion/pcf-zerodowntimedeploy). The main reason is the fact that we use plugins for Cloud Foundry in Jenkins and Urban Code Deploy, which only support the push
command and does not allow to use zero-downtime-push
instead.
@adamsiemion Thanks for your support. Hopefully @xoebus will find the time to take a look.
I have created a new arg flag
keep-old-app
, which skips deleting of the old app.This was required by the project which currently work on. We use Pivotal Cloud Foundry (PCF) container-to-container networking and Eureka service registry.
The way we use
autopilot
is the following (let's assume the service we are redeploying is calledtest-service
:test-service
already registeredzero-downtime-push
with-keep-old-app
test-service
(the new app) andtest-service-venerable
(the old app) are both running, but the new app is not registered in Eureka because it cannot establish a connection to Eureka due to the missing c2c policies therefore all the requests are still routed to the old servicecf delete test-service-venerable
to remove the old app