More questions. It is generally about how to organize in a large organization. If I manage to get acceptance for pulumi I imagine something along these lines: 1. One project that deals with all infrastructure 2. One project per app that can reference the infra stack to get access to container registry I have something basic working along those lines, but have some questions. Let's say that I in step 2 would like to build container image and push it to a container registry defined in step 1, and have that run on the CI server after build. This should not deploy the app, just push it to the registry. Later when deploying the app I want to use the image created. This is still all good, the question is:

Can I have one project that does the build image and also deploys based on an input flag when running

pulumi up

, or should I have one project for package the app and one for deploy?

I have a situation in a TS project where, when I delete an unnecessary import, Pulumi wants to delete 7/8 of my resources. The line is

-import * as kx from "@pulumi/kubernetesx";

It's a change to my

index.ts

. Has anyone else experienced this kind of thing? Something similar happened a couple of days ago when I reordered my imports but I didn't stop to investigate.

Hi, I am trying to get an ECS stack up using crosswalk , Is it possible to configure dynamic port Mapping using crosswalk ? Below is the code for creating an Ec2Service

const service = new awsx.ecs.EC2Service(name, {
    name: `${name}-${utils.resolveStack()}`,
    cluster: cluster,
    subnets: ["xxxxx"],
    taskDefinitionArgs: {
      networkMode: "bridge",
      taskRole: ecsConfig["taskRole"],
      logGroup: null,
      executionRole: ecsConfig["executionRole"],
      containers: {
        test: {
          image: "xxxxx",
          cpu: 1024,
          memory: 1024,
          portMappings: [listener]
        }
      }
    },
    desiredCount: 0
  });

The above config creates the Service with a loadbalancer/target group that is configured over port 8080 , this forces a new target to be registered to the target group over 8080 only , But in my case I want the targets to register over ephemeral ports If I want multiple docker containers to run on the same container instance , I need dynamic mapping which is configured via

portMappings:[
{
hostPort:0,
containerPort:8080
}]

If I do the above , the Service is created without a load balancer , Is there a way to get this to work per https://aws.amazon.com/premiumsupport/knowledge-center/dynamic-port-mapping-ecs/ Any help is appreciated 🙂

Are releases announced here in slack?

Only "big" releases generally, like 2.0 or other major feature releases and in #announcements as Scott said.

howdy. been trying pulumi to shift away from terraform, however pulumi's output type handling is quite the challenge to wrap my head around. i've been reviewing months of slack conversations, reading docs, searching the web, yet i'm not sure how to proceed with it. i'm using typescript since that seems to be the best supported language (examples, docs, and crosswalk all either only support TS or favor it by tending to have more information in this language). i'd like to loop over the list of ASG names in this situation, yet it clearly iterates on the promise object because the plan shows

__pulumiOutput

and similar as values for

g

. preview:

pulumi:pulumi:Stack        gcso-k8-dev                
 +   ├─ aws:ec2:LaunchTemplate  __pulumiOutput  create     
 +   ├─ aws:ec2:LaunchTemplate  promise         create     
 +   ├─ aws:ec2:LaunchTemplate  toString        create     
 +   ├─ aws:ec2:LaunchTemplate  toJSON          create     
 +   ├─ aws:ec2:LaunchTemplate  allResources    create     
 +   ├─ aws:ec2:LaunchTemplate  resources       create     
 +   ├─ aws:ec2:LaunchTemplate  isSecret        create     
 +   └─ aws:ec2:LaunchTemplate  isKnown         create

what am i not grasping?

is pulumi able to create resources based upon the query results of other resources being created in the same

up

operation? granted that leaves preview rather useless at predicting what resources would be CRUD'ed, but is it possible?

Any examples out there of using the digital ocean container registry? This seems to be missing the crucial bit: https://www.pulumi.com/docs/reference/pkg/nodejs/pulumi/digitalocean/#docker-provider-example

Hey I’ve got a project organization question. i was hoping to have a single project with a stack per component (there’d actually be three stacks per component: <component>-dev,<component>-stage, <component>-prod. I was hoping to push the Pulumi stack files into each component:

project_root
├── Pulumi.yaml
├── README.md
├── elasticsearch/
├── fargate/
├── go.mod
├── go.sum
├── influxdb/
├── main.go
├── main_test.go
└── s3sink/
    ├── Pulumi.sink-dev.yaml
    ├── Pulumi.sink-stage.yaml
    ├── Pulumi.sink-prod.yaml
    ├── handler/
    ├── handler.zip
    ├── policies.go
    ├── sink_bucket.go
    └── sink_bucket_test.go

When I try running

pulumi up

I get an error saying config hasn’t been set. I was wondering if anyone has any pointers (aside from moving the stack yaml files to the same level as the project)…

Hi, Is someone able to have a look at my message? https://pulumi-community.slack.com/archives/C84L4E3N1/p1591513298487900 Thank you!

Is there a way to wait for Kubernetes objects (resources) to become available? Let's assume I deploy a Kubernetes Operator using Pulumi and that operator will then register some CustomResourceDefinitions (CRD) for which I want to create a CustomResource instance using Pulumi. From my Pulumi's stack point of view there is currently no way to figure out whether the CRDs do already exists. 1. Deploy an Operator using Pulumi (e.g. as Deployment or via Helm) 2. Wait for the CRDs to be created by the operator 3. Create instances CR instances for that CRD using Pulumi As of today I revert to a dynamic resource which will block (sleep) a certain amount, but that approach is error prone and slow.

congratulations to pulumi on securing the reddit.com/r/pulumi subreddit

Hey folks, is there any way to build Dynamic Provider using .NET Core(c# f#)?

hey all! im trying to make a small prototype of a data pipeline using pulumi, aws step functions, and aws lambda where the lambda function also relies on some python dependencies (like pandas). im working off of this example: https://github.com/pulumi/examples/tree/master/aws-py-stepfunctions but it doesn’t involve packaging any 3rd party python libraries. can anyone point me to more documentation or a more detailed example on how to accomplish this? thanks 🙏

any further documentation on https://github.com/pulumi/pulumi-tf-provider-boilerplate ?

Is there a good way to do something like this: https://hasura.io/docs/1.0/graphql/manual/deployment/kubernetes/index.html#deploy-kubernetes

I want to use Typescript to pull those YAML files, then update the URL in the one field with a URL of a resource I create via Pulumi and apply them

how do people generally manage common configuration in multiple accounts? eg. roles, cloudtrail config etc that you want to be standard across the org? Do you create a “core-infra” project with a single stack and then just loop over accounts within it? Create multiple stacks?

Hi everyone, I was wondering if there's any example of deploying resources with multiple instances (eg, an Azure App Service with multiple instances in multiple locations), with instances depending on each other? For example, for Azure, you need to create an App Service Plan for App Services; I was wondering if there's an example I can checkout that creates multiple plans and then uses those plans to create multiple apps. I want to see how naming is handled in these situations

pulumi tries to use wrong version for gcp plugin, seems like a bug to me:

$ pulumi up
...
Diagnostics:
  pulumi:providers:gcp (default_2_1_1):
    error: no resource plugin 'gcp-v2.1.1' found in the workspace or on your $PATH, install the plugin using `pulumi plugin install resource gcp v2.1.1`

Version number comes from plugin

random

- I can prove this by upgrading it:

$ yarn add @pulumi/random@2.1.2

$ pulumi up
...
Diagnostics:
  pulumi:providers:gcp (default_2_1_2):
    error: no resource plugin 'gcp-v2.1.2' found in the workspace or on your $PATH, install the plugin using `pulumi plugin install resource gcp v2.1.2`

this is pulumi 2.4.0 btw

I'm not sure how to fix this, any help is appreciated

update: reverted back changes in package.json, fixed now.

Pulumi import functionality really hurts. Does anyone has solid way to import k8s resources which are deployed from Helm packages? Do I have to write specific transformations for this?

I am deploying operator-like application with CRDs and till the recent moment pulumi/kubernetes package was not able to create them, so I added them manually in the beginning. Now I upgraded pulumi/kubernetes and it wants to create CRDs but they already exist...

this is what I came up with: 1. export stack state with

pulumi stack export > stack.json

2. manually add CRD into

stack.json

3. import state back with

pulumi stack import --file stack.json

4. refresh resource state with

pulumi refresh -t 'urn:...'

- this is to prevent replacement of CRD 5. sync with regular

pulumi up

but this is rather tedious and I think it'll be useful to have some pulumi subcommand for this.

Hello, just learning pulumi, I have a use case of attaching a AWS lambda function to multiple AWS sns topics. Following the examples I was able to build this up relatively quickly by looping over the topics names and attaching event handlers. I noticed that the pulumi runtime created separate lambda functions for each topic even though the function is exactly the same for all topics. Is there a way to attach multiple triggers to a lambda function, or just reuse a lambda function. I think manually creating the lambda function as a resource maybe an option but would love to know if there's a shorter way.

One common topic when promoting Pulumi is the declarative vs imperative debate. Many people seem to prefer declarative in infrastructure because of the maintainability aspect. My understanding is Pulumi is also declarative, but the more dynamic nature of real programming languages seems to scare a bit some Terraform users. What are some of the techniques people here use to have more predictability in the infrastructure they are building? I was thinking about using Starlark + Mypy instead of normal Python for example. I would love to hear other thoughts.

I have just tried pulumi for the first time using pycharm against AWS (i have created s3 buckets etc) it all seems straight forward. Is there a way to debug the python code interactively using the pycharm ide?