I'm trying to run my application on Kubernetes. My docker container have environment variables such as PATH and LD_LIBRARY_PATH, which are set in Dockerfile. I tried to change them in the yaml file like this:
env:
- name: LD_LIBRARY_PATH
value: "foo:$(LD_LIBRARY_PATH)"
The above configuration doesn't work, I just see LD_LIBRARY_PATH=foo:$(LD_LIBRARY_PATH) in the pod. This method seems work for Kubernetes env variables such as KUBERNETES_PORT_443_TCP_PROTO, but not for docker env variables.
My questions are:
I think the env settings in yaml are injected into docker before the running time of the container, so kubernetes cannot read the value of LD_LIBRARY_PATH. Therefor it can't change the variable. Do I understand it right?
How to change container environment variables with kubernetes env? I know that I can set the env variables in the command field of yaml file, but that seems not clean and are there other ways to do that?
If Kubernetes can't change existed envs, does it mean that the env field in yaml file is designed to add new envs only?
Thank you!
The Kubernetes variable expansion syntax only works on things Kubernetes directly knows about. Inside a container an environment variable could come from a couple of places (the Dockerfile ENV directive, the base container environment itself, setup in an entrypoint script) and Kubernetes doesn't consider any of these; it only considers things in the same container spec. The API definition of EnvVar hints at this:
Variable references $(VAR_NAME) are expanded using the previous defined environment variables in the container and any service environment variables.
You can't use this Kubernetes syntax to change environment variables in the way you're describing. You can only refer to other things in the same env: block (which may come from ConfigMaps or Secrets) and the implicit variables that come from other Services.
(Changing path-type variables at the Kubernetes level doesn't make a lot of sense. Since an image is self-contained, it already contains all of the commands and libraries it would need. It's difficult in Kubernetes to inject more tools or libraries; it'd be better to install them directly in your image, ideally in /usr/lib or /usr/local/lib, but failing that you can update ENV in a Dockerfile similar to how you suggest here.)
Related
How i can access or read windows environment variable in kubernetes. I achieved the same in docker compose file.
How i can do the same in kubernetes as i am unable to read the windows environment variables?
Nothing in the standard Kubernetes ecosystem can be configured using host environment variables.
If you're using the core kubectl tool, the YAML files you'd feed into kubectl apply are self-contained manifests; they cannot depend on host files or environment variables. This can be wrapped in a second tool, Kustomize, which can apply some modifications, but that explicitly does not support host environment variables. Helm lets you build Kubernetes manifests using a templating language, but that also specifically does not use host environment variables.
You'd need to somehow inject the environment variable value into one of these deployment systems. With all three of these tools, you could include those in a file (a Kubernetes YAML manifest, a Kustomize overlay, a Helm values file) that could be checked into source control; you may also be able to retrieve these values from some sort of external storage. But just relaying host environment variables into a container isn't an option in Kubernetes.
I'm a beginner in NiFi setup. I'm planning to start a NiFi cluster on Kubernetes. In normal installation, I saw that, we can change the NiFi configurations under the file 'nifi.properties'. But, when it comes to docker image, I also saw that we can change that by using environment variables. In most of the cases, the properties mentioned in the nifi.properties file can be easily converted into its equivalent environment variable.
Eg:
nifi.web.http.host <=> NIFI_WEB_HTTP_HOST
But in some cases, the environment variable is different. Eg:
nifi.zookeeper.connect.string != NIFI_ZK_CONNECT_STRING
From where do we get the full list of NiFi environment variable for Docker image. Any help like links or directions is very much appreciated.
You need to look into the documentation (or the source code) of the NiFi docker images your are using. For example agturley/nifi and apache/nifi.
When you enter the docker container you can see secure.sh and start.sh under the path /opt/nifi/scripts. These are the scripts that make all prop_replace
I need to change an environment variable in my deployment (nothing else has changed). When I update my environment variables (just using export) and use docker stack deploy, the container environment is not updated. How do I make sure environment variables are updated in the container?
Apparently using docker stack deploy should work to update the environment. My error was not having the environment variable defined in the docker-compose.yml file.
I am deploying some apps in kubernetes,and my apps using a config management tool called apollo.This tool need to define the apps running environment(develop\test\production......) through this ways:1.java args 2.application.properties 3./etc/settings/data.properties. Now I am running apps in Kubernetes,the question is,how to define running environment variable?
1.if I choose java args,so I should keep some scripts like: start-develop-env.sh/start-test-env.sh/start-pro-env.sh
2.if I choose application.properties,I should keep application-develop.properties/application-test.properties.....
3.if I choose /etc/settings/data.properties,It is impossible to login every docker container to define the config file of each environment.
what is the best way to solve the problem? write in kubernetes deployment yaml and my apps could not read it(define variable in batch pods collections in one place is better).
You can implement #2 and #3 using a configmap. You can define the properties file as a configmap, and mount that into the containers, either as application.properties or data.properties. The relevant section in k8s docs is:
https://kubernetes.io/docs/tasks/configure-pod-container/configure-pod-configmap/
Using java args might be more involved. You can define a script as you said, and run that script to setup the environment for the container. You can store that script as a ConfigMap as well. Or, you can define individual environment variables in your deployment yaml, define a ConfigMap containing properties, and populate those environment variables from the configmap. The above section also describes how to setup environment variables from a configmap.
I would like to be able to test my docker application on local before sending it to the cluster. I want to use mini Kube for this. Meanwhile, instead of having multiple kube config files which would define env variables for the cloud environment and for my local machine, I would like to override some of the env variables when running in local. I can see that you can do something like that with docker compose:
docker-compose up -f docker-compose.yml -f docker-compose.e2e.yml.
The second file would only have the overriding values. Yes, there are two files but I find it clean.
Is there a way to do something similar with Kube/minikube? Or even something better ???
I think you are asking how to pass different environment values into your Pods depending upon which environment they are deployed to. One pattern to achieve this is to deploy with helm. Then you use templated versions of your kubernetes descriptors for deployment. You also have a values.yaml file that contains values to be injected into the descriptors. You can switch and overlay values.yaml files at the time of install to control which values are injected for a given installation.
If you are asking how to switch whether a kubectl command runs against local or cloud without having to keep switching your kubeconfig file then you can add both contexts to your kubeconfig and use kubectl context to switch between them, as #Ijaz Khan suggests