Hello, We'd like to set up a scheduled cleanup of our Artifact Registry but we don't want that to affect release images. The idea that would work best with our tool and also sounds like a general good practice would be to tag release images with a release tag. Unfortunately, I can't find a non-hacky way to do that. I tried to fetch info about images garden builds in a given release job - seems like "v-*" tag would be the most reliable - and then use that to locate image in registry to add the release tag using gcloud but none of the solutions is satisfactory. Here's what I tried: - Add

--output=json

flag and grep "v-.\*" on the output - the problem with this solution is that we prefer to see regular logs in pipeline for debugging and monitoring progress and in this way we'd only see json. - Fetch info from one of the logs in

.garden/logs

- the problem here is that grepping gets even more complex and it's relying on the log structure and wording to stay the same which doesn't seem reliable and would be difficult to track Is there any other method you could recommend?

I have a kubernetes module (

myservice

) which depends on an image that is built from a container module (

myimage

).

myservice

has

myimage

in it's build dependencies When I

garden deploy myservice

, it kicks off the build

myimage

and deploy

myservice

steps at the same time. Often,

myservice

will timeout while waiting for the build to finish. Is it possible to configure garden to not even start to deploy

myservice

until

myimage

is done building?

Datadog uses labels for lots of things - I have been unable to figure out how to add labels to services deployed via Garden - is there a way?

Hello, When we change git branches, and try to deploy something new, we always get an error message that the ephemeral namespace doesn't exist and we have to create it. To clarify my question, consider these steps. 1. We are in repo X with branch Y. We create a new environment, suppose this new environment has the namespace: Y 2. We deploy and everything is good and happy. 3. Switch branch, or even just create a new branch based off the current branch. 4. Now if we try to deploy, Garden complains that namespace

Z

does not exist. It does not keep using the namespace

Y

that was already established and deployed to. Please note that

.garden

is ignored and not part of the git repo, so no issues there. Alternatively is there a garden cli command to force the usage of a namespace? Thank you.

Hey nice garden folks! I would like to use a separate docker image when running in dev mode, how can I achieve that in the most straightforward way? I couldn't find

Is there a way to cast a double to to an int...? My usecase is to properly size jvm heap..:

yaml
# module..
variables:
  memory-mb: 512

# k8s manifest
env:
  - name: HEAP_OPTS
    value: "-Xms${(var.memory-mb * 0.9) / 2}m -Xmx${var.memory-mb * 0.9}m"

Unfortunately, jvm doesn't take doubles, it wants ints. The only way I could come up to solve this with the existing templating functions is this one:

split(string((var.memory-mb * 0.9) / 2), '.')[0]

->

Unable to parse as valid template string.

It doesn't work due to https://github.com/garden-io/garden/issues/3001. A

toInt

function would really come in handy here. Am I missing something?

I would like to modify some aspects of the podspec but I only see

nodeSelector

exposed.

Hey there! Here are some recommendations to make a greater impact with your forum posts. First, before posting make sure you have read the following.: 1. Look for any other similar issues in this forum #1044642930075373588 if you find your issue there, make sure to follow the thread and feel free to add your experiences to it. 2. If you can try to include snippets or reproducible examples. They're a huge help to solving your issue. 3. Be friendly! 👧 🤝 👦 Example 🆘 Post Hi everyone, I'm experiencing [insert issue/problem here]. [Insert any additional details about the problem or issue you're facing.] I've tried [insert any steps you've already taken to solve the problem]

I get the following message when doing a cluster init if I use the garden nginx ingress controller: "container controller specifies host ports [80 443], which are disallowed in Autopilot." I have tried following the kubernettes.io instructions for using nginx ingress, setting it up as a kubernetes module, but that requires adding custom args to the deployment - which I can't find a way to do. I can create the service this way, but it appears to be missing a link to the exposed services config map as its not actually forwarding requests and in the gke console it shows no mapped (https://kubernetes.github.io/ingress-nginx/user-guide/exposing-tcp-udp-services/) Do I need to manage the load balancer/ingress directly with kubectl? Is there some other magic at play here?

I have a project for which it would be a pain to convert all of my services using the container module type to k8s module type simply to specify which nodes the pods are scheduled on. Is there any exposure of tolerations/affinity on the container module type or any plans for this? Is there an alternative way to do this?

Is there an easy way to debug a) what variables are available and b) what is the value (recursively) of a variable in a given template?

Given I have a fresh k8s and no .garden folder, when I do a

garden deploy

where I have a kubernetes or a container module, that defines ingresses, then the module fails because it tries to create the ingress too early:

Failed deploying service 'prometheus' (from module 'prometheus'). Here is the output:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Command "/Users/xy/.garden/tools/kubectl/f8a39834529cd1ed/kubectl --context=minikube apply --output=json -f -" failed with code 1:

Error from server (InternalError): error when creating "STDIN": Internal error occurred: failed calling webhook "validate.nginx.ingress.kubernetes.io": failed to call webhook: Post
"https://ingress-nginx-controller-admission.ingress-nginx.svc:443/networking/v1/ingresses?timeout=10s": dial tcp 10.101.27.41:443: connect: connection refused

Error from server (InternalError): error when creating "STDIN": Internal error occurred: failed calling webhook "validate.nginx.ingress.kubernetes.io": failed to call webhook: Post
"https://ingress-nginx-controller-admission.ingress-nginx.svc:443/networking/v1/ingresses?timeout=10s": dial tcp 10.101.27.41:443: connect: connection refused

It succeeds when I rerun

garden deploy

I'm trying to create HorizontalPodAutoscaler using $forEach to generate all the manifests and keep it DRY but I don't see a way to get an array of all the module names in the template string docs. It's not the first time this would be useful and usually I work around it but this time I don't see anything that wouldn't rely on just listing the modules by hand and then having to maintain that list up to date.

Hey all, how do I uninstall garden.io completely on MacOS? Simply running

brew remove garden-cli

still leaves garden running (tried going to http://localhost/something and got the garden 404 page still that says it's running)

Hello! I'm currently building a dev cluster of sorts where I want myself a few others to be able to deploy their code to run in isolated namespaces. I'm pretty much done with everything, but one problem I have is that some resources, like

ExternalDNS

, are not specific to a particular developer's namespace. What would be the recommended way to manage deployments that are done at the cluster level vs deployments done for a particular namespace. (I know my wording is confusing - i'm not really sure the best way to phrase the question 😛 )

Hi all, is there any guide how to optimize Kaniko in cluster build. We currently have the problem that we run into timeout with yarn builds for the react frontend because the CPU is exhausted. How can i scale Kanikopods independently via garden.

How to ask Garden to deploy a service in dev mode and its dependencies only vs. entire stack defined in garden.yml file? For example, my master garden file has configuration for all projects within our organization. I don't want all services to get deployed in K8. I want the service I'm working on being deployed in dev mode with its dependencies in regular deployed mode. Looking at the examples https://docs.garden.io/reference/commands#garden-deploy its not clear.

Hello, I am trying to use ModuleTemplate. While trying to deploy the service, I am unable to add

buildArgs

to the

templated

module. I get the error

key "buildArgs" is not allowed at path [buildArgs]

If I try to add

buildArgs

in the template itself and use

${inputs.XXX}

it complains

Could not find key inputs. Available keys: command, git, local, project and secrets.

Basically it looks like inside the template I can not use

inputs

except in the

modules

section. So currently as it stands I can not use

buildArgs

with

templates

. This is a blocker for using this feature. Is there any other way around it?

Wondering if it is possible to cache exec modules? For my stack I first run a build (not a task, so I can run tests on the build) to generate my executables that I copy into my containers. Currently using exec reruns this every time. I understand that for the other modules the caching behavior is baked in and there is no simple way to cache exec calls. Maybe I have to wait for the plugin SDK?

Can we define a new

ModuleTemplate

based off an existing one? The idea here is the project is giving, same template X and then a set of modules can create their own template Y based of template X and define the arguments that are common for all of it's services instead of repeating them for each template of the base template X. Can this be done? I tried an example but I can't get it to work

Hello, guys! So the problem is that i cant access my services within cluster with HTTP-requests. For example, i have two services:

frontend and api-gateway

. Their ports are respectively 3000 and 3001. The main issue is that i cant do a HTTP-request to http://api-gateway:3001 ( i get an

ERR_NAME_NOT_RESOLVED

). I can only do that using

external IP-address (http://127.10.0.7:3001)

. When i used Docker Compose, i had no problems with getting access to the other services with

api-gateway:3001

. I am sorry in advance for it, but can i ask one more question? I run my own test in Garden, but i cant see the

succeed

option in the Dashboard near it. It uses Jest-library to run the test. And it really passes it. Thanks in advance!

Hi, is there any planned work / update for garden to start supporting OCI helm repos? Currently it completely does not work, and bitnami is (slowly but surely) updating their charts to reference OCI charts as dependencies, thus breaking gardens ability to deploy them. The workaround is to simply pin the chart to the version before the dependency upgrade, but even that will eventually stop working once they stop the (old) helm index. e.g. try deploying this without the pinned version

yaml
kind: Module
type: helm
name: kafka
include: []
repo: https://charts.bitnami.com/bitnami
chart: kafka
# version: 21.4.6
values:
  maxMessageBytes: "10000000"
  persistence:
    enabled: false
  numPartitions: 10

Is there a reason why workflows can't have the command

run workflow XXX

We have one workflow specifically designed for BuildKite and it could use calling other workflow instead of having to duplicate the steps there and copying them over... I think we can all agree copying is bad, will break the single source of truth concept. Thank you.

It seems in my current setup

.garden/error.log

is always empty. However I find the actual file under

.garden/logs/error.log

Alongside

development.log

How can I fix this to not go under logs folder and just stay on the root folder. I searched the documents and I could not find a way to influence this via an env variable or some sort of config. Thank you!

Maybe this would be better as an issue on github, but it is kind of unrelated to the codebase itself. Anyways, I am starting to use the gardendev/garden-gcloud container for running garden in CI, and I've found that the gcloud application is a bit outdated even in the latest tag. Would be awesome if you could publish a version with a more up to date version. Here is the message I received when using this image:

gardendev/garden-gcloud:latest

$ gcloud --quiet components update
Beginning update. This process may take several minutes.
Your current Google Cloud CLI version is: 411.0.0
You will be upgraded to version: 428.0.0
┌─────────────────────────────────────────────────────────┐
│            These components will be updated.            │
├─────────────────────────────────┬────────────┬──────────┤
│               Name              │  Version   │   Size   │
├─────────────────────────────────┼────────────┼──────────┤
│ BigQuery Command Line Tool      │     2.0.91 │  1.6 MiB │
│ Bundled Python 3.9              │     3.9.16 │ 63.4 MiB │
│ Cloud Storage Command Line Tool │       5.23 │ 15.5 MiB │
│ Google Cloud CLI Core Libraries │ 2023.04.25 │ 20.3 MiB │
│ Google Cloud CRC32C Hash Tool   │      1.0.0 │  1.2 MiB │
│ anthoscli                       │     0.2.36 │ 68.4 MiB │
│ gcloud cli dependencies         │ 2023.04.25 │ 11.3 MiB │
│ gke-gcloud-auth-plugin          │      0.5.2 │  7.7 MiB │
│ kubectl                         │    1.24.13 │  < 1 MiB │
│ kubectl                         │    1.24.13 │ 92.6 MiB │
└─────────────────────────────────┴────────────┴──────────┘
A lot has changed since your last upgrade.  For the latest full release notes,

and so on.

Hi Garden team! We recently upgraded our CI cluster to Kubernetes 1.23, and after we upgraded, we began to see intermittent "socket hang up" errors printed from Garden, followed by an immediate termination. I'll post an example log in the thread. After doing a lot of troubleshooting and searching, we think that upgrading to Node 19+ might fix the issue. It makes some changes to networking that might fix the problem. Could ask for a favor - could you publish a pre-release build of Garden that uses Node 19 or Node 20, so we can verify? Thank you!

Is it possible to have ingress like functionality but to non ingress endpoints? (TCP/UDP) for DB and other services?

One of my coworkers has been getting the following error when running garden dev and cannot use his dev environment because of it.

/snapshot/project/tmp/pkg/cli/node_modules/execa/lib/error.js:59
        error = new Error(message);
          ^
Error: Command "git status" failed with code 128:
fatal: this operation must be run in a work tree
    at makeError (/snapshot/project/tmp/pkg/cli/node_modules/execa/lib/error.js:59:11)
    at handlePromise (/snapshot/project/tmp/pkg/cli/node_modules/execa/index.js:114:26)
    at processTicksAndRejections (node:internal/process/task_queues:95:5)
    at exec (/snapshot/project/tmp/pkg/cli/node_modules/@garden-io/core/src/util/util.ts:231:17)
    at /snapshot/project/tmp/pkg/cli/node_modules/@garden-io/core/src/vcs/git.ts:97:26
    at wrapped.ensureSafeDirGitRepo (/snapshot/project/tmp/pkg/cli/node_modules/@garden-io/core/src/vcs/git.ts:162:7)
    at wrapped.getRepoRoot (/snapshot/project/tmp/pkg/cli/node_modules/@garden-io/core/src/vcs/git.ts:205:5)
    at wrapped.getFiles (/snapshot/project/tmp/pkg/cli/node_modules/@garden-io/core/src/vcs/git.ts:269:21)
    at findConfigPathsInPath (/snapshot/project/tmp/pkg/cli/node_modules/@garden-io/core/src/util/fs.ts:175:17)
    at /snapshot/project/tmp/pkg/cli/node_modules/@garden-io/core/src/watch.ts:298:29

Here is more context about our setup and about the encoutered problem: - Garden version: 0.12.55 - We include 3 remote sources in the garden.project.yaml file. - We have a deployment bash script which runs

garden link source

on those remote sources and then

garden dev

- The problem is sadly only reproducible on my coworkers MacOS laptop at the moment. I would appreciate any leads on this since it's affecting my coworker's productivity by quite a lot. 😦

Hello, Has anyone had a chance to look into: https://github.com/garden-io/garden/issues/4174

Hello, We have a task that we run on a database image that exports some artifacts. While the task is doing it's thing we get a lot of these

outputs

. It looks like it is not affecting functionality

MaxListenersExceededWarning: Possible EventEmitter memory leak detected. 11 drain listeners added to [ReadEntry]. Use emitter.setMaxListeners() to increase limit

We get about 30 - 50 of such outputs. Please find attached image.