This message was deleted.
# generals
sparse-intern-71089
02/11/2021, 6:28 AMThis message was deleted.
a
ambitious-father-68746
02/11/2021, 10:45 AMRefresh does't change any resources, it simply updates the state file with what's on the real world.
ambitious-father-68746
02/11/2021, 10:46 AMFor example, imagine you have an EC2 instante with instance class t2.nano, started with Pulumi. Then you go to the EC2 console and upgrade it to m5.xlarge. Pulumi doesn't know about this change, so it won't try to bring it back to t2.nano.
ambitious-father-68746
02/11/2021, 10:48 AMIf you run refresh, Pulumi will look at AWS, see that the instance is now m5.xlarge and it will update the state file accordingly. Next time you run Pulumi it will know that it needs to change the instance class of the instance back to t2.nano.
b
blue-orange-84367
02/11/2021, 10:54 AMthanks for the answer.
blue-orange-84367
02/11/2021, 10:54 AMis it then an antipattern to run refresh in a ci pipeline?
blue-orange-84367
02/11/2021, 10:55 AMe.g i ran into the issue that when changing my access key for my aws account, the refresh fails (as the key has changed which makes sense) and the next
pulumi upa
ambitious-father-68746
02/11/2021, 11:53 AMIf refresh messes with your state file, then that might happen
b
bored-oyster-3147
02/11/2021, 2:17 PM@blue-orange-84367 maybe consider doing something else to test your access key before running pulumi refresh. Though I'm surprised that pulumi refresh failing would update the state file
b
brave-angle-33257
02/11/2021, 4:23 PMi re-run refresh on all my ci/cd, just too high of a chance that something has changed and it might break the deploy. some resources don't like updates as well as others, or you don't want cascading deletion of other resources, so in those cases I'll delete a resource from a console, update my git repo and run a deploy, it will see the item is gone, and recreate it with new configuration.
b
blue-orange-84367
02/12/2021, 5:06 AMactually the refresh isn't failing. It prints the following error in the build logs:
Copy code
refreshing warning: configured Kubernetes cluster is unreachable: unable to load schema information from the API server: the server has asked for the client to provide credentialblue-orange-84367
02/12/2021, 5:08 AMIt seems like afterwards the state lost all knowledge about the deployment in the cluster and tries to create new deployments.
as said we disabled auto-naming (for some internal reasons), which leads to pulumi complaining, that a resource with that
name already exists.