pulumi-community #general

general

busy-soccer-65968

06/12/2020, 5:29 PM

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

enough-kite-69616

06/12/2020, 6:04 PM

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

enough-kite-69616

06/12/2020, 6:05 PM

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

important-appointment-55126

06/12/2020, 9:55 PM

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?

rich-napkin-40911

06/13/2020, 12:03 AM

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

elegant-dress-88912

06/13/2020, 2:21 AM

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`

elegant-dress-88912

06/13/2020, 2:22 AM

this is pulumi 2.4.0 btw

elegant-dress-88912

06/13/2020, 2:23 AM

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

elegant-dress-88912

06/13/2020, 2:26 AM

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

elegant-dress-88912

06/13/2020, 2:32 AM

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?

elegant-dress-88912

06/13/2020, 2:34 AM

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...

elegant-dress-88912

06/13/2020, 2:56 AM

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.

bitter-australia-87528

06/13/2020, 4:32 AM

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.

orange-fountain-17148

06/13/2020, 10:23 AM

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.

bored-finland-61499

06/13/2020, 3:54 PM

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?

cold-motherboard-287

06/13/2020, 4:23 PM

Let's say I want to create a new provider (I guess that's how it's called) for another cloud provider.... Where could I start? do you guys use a specific language (from the ones you support) in the core or could I use any for such a task?

mysterious-pilot-36551

06/14/2020, 12:38 AM

Is the

depends_on

option supposed to delay the creation of a resource until the resource it depends on finishes being created? I have a Kubernetes custom resource that depends on a Chart, however, Pulumi is trying to create both the Chart and the custom resource at the same time. This causes it to fail because the CRD isn’t created until the Chart finishes getting created. Here is an excerpt. Chart is declared on line 26 and the custom resource on line 46.

mysterious-pilot-36551

06/14/2020, 12:44 AM

Nevermind, this is a known bug: https://github.com/pulumi/pulumi-kubernetes/issues/861

numerous-lifeguard-13411

06/14/2020, 2:09 AM

Hey folks, I'm trying to do something that seems like it should be simple: set up a no-scaling, no-redundancy bastion-web-app configuration. But I'm getting an infuriating pair of errors, one saying it's expecting a security group to start with "sg-" for its name, and another that says "sg-" is a forbidden patter for security groups.

numerous-lifeguard-13411

06/14/2020, 2:10 AM

aws:ec2:SecurityGroup (appserver):
    error: 1 error occurred:
    	* creating urn:pulumi:dev::cisco-exercise::aws:ec2/securityGroup:SecurityGroup::appserver: Error authorizing security group ingress rules: InvalidGroupId.Malformed: Invalid id: "bastion-05865a3" (expecting "sg-...")

numerous-lifeguard-13411

06/14/2020, 2:10 AM

Any clues there?

numerous-lifeguard-13411

06/14/2020, 2:31 AM

Here's the other version:

aws:ec2:SecurityGroup (sg-bastion):
    error: Error creating Security Group: InvalidParameterValue: Value (sg-bastion-1f46895) for parameter GroupName is invalid. Group names may not be in the format sg-*.

numerous-lifeguard-13411

06/14/2020, 2:31 AM

You can see why this might make a fellow doubt his own sanity.

numerous-lifeguard-13411

06/14/2020, 3:18 AM

Seems like for security groups, the

tags: { Name: "sg_blah" }

is required

little-garage-43399

06/14/2020, 11:12 AM

I think there is a import bug in Pulumi.

from pxglb import *

pxglb is a file with classes I have built. This import is worked fine, until I have added to pxflb file a new class, a "ResourceProvider" class

[1:05 PM] class PXNS1RecordProvider(ResourceProvider):

In this case I get the following error:

error: Exception calling application: No module named 'pxglb'

If I copy paste the classes code into the "main" file (instead of the import), it works fine. Pulumi version: 2.4.0 Python version: 3.7.7

famous-sugar-52038

06/14/2020, 4:51 PM

Hello, there 🙋‍♂️ I have been wasting a lot of time to update k8s clusters… As I know that pulumi has superpowers to deploy with general-purpose languages, what if I used pulumi to update k8s clusters. Is it more simpler than using a plane k8s or managed-k8s with ymls? 🤔

numerous-lifeguard-13411

06/14/2020, 5:29 PM

OK, dumb question, how do I specify a local route on a routeTable?

numerous-lifeguard-13411

06/14/2020, 11:09 PM

I have fought through many trials. Now I have one last stupid requirement to deal with: hosts must not have Internet access. But they also are supposed to have httpd and curl installed on them. I could do this by giving them access in the initial configuration, letting them pull their installs, and then editing the config to remove the egress rules, and then doing

pulumi up

again. But that seems pretty grossly manual.

numerous-lifeguard-13411

06/14/2020, 11:09 PM

Is there a way to do a "staged" deploy, of sorts?

average-receptionist-65239

06/14/2020, 11:55 PM

pulumi up

command is "Waiting for app ReplicaSet be marked available (0/1 Pods available)". However, the deployment is ready. The pod is operating just fine.

Title

average-receptionist-65239

06/14/2020, 11:55 PM

pulumi up

command is "Waiting for app ReplicaSet be marked available (0/1 Pods available)". However, the deployment is ready. The pod is operating just fine.

Timed out with

[1/2] Waiting for app ReplicaSet be marked available (0/1 Pods available)
error: 2 errors occurred:
	* the Kubernetes API server reported that "default/search-indexer" failed to fully initialize or become live: 'search-indexer' timed out waiting to be Ready
	* Minimum number of Pods to consider the application live was not attained

This is the third time I've tried this operation. I appear to be stuck. Anyone know a workaround?

I deleted the deployment with

kubectl

and did a

pulumi up --refresh

to work around the issue.

I noticed a couple of closed bugs that seem relevant https://github.com/pulumi/pulumi-kubernetes/issues/485 https://github.com/pulumi/pulumi-kubernetes/issues/497

View count: 1