How to rollback Istio to the old control plane, instead of completing the canary upgrade?
We were doing a test on Canary upgrade of our Istio setup from 1.7.3 to 1.8.6 version.
So we installed canary version 1-8-6 for the istio controlplane and we mapped one namespace to the upgraded version. We tested the application and we were getting 200 response.
Everything was working as expected till this point. So in order to try a rollback, we just went to the document but the steps are not clear enough regarding the rollback.
What is the method to reinstall the gateway here? We used istioctl with overlay and default profile for v1.7.3 installation. So, how to reinstall the gatway?
After uninstalling the canary version, the ingress gateway pods got removed and our application got down with a 500 error.
So would like to clear about the exact process to rollback from canary version to previous installed version. But due to some issues, we want to rollback to previous version.
You should not do
istioctl x uninstall --revision=canary
if you successfully upgraded. This command should only be used if you decided to stop upgrading in the middle of the process, and want to go back.
To downgrade to a lower version of Istio you should use
istioctl upgrade. Steps are identical to actual upgrade procedure, you just have to use istioctl binary corresponding to a lower version.
You can find more in the docs.
If you still want to go with istioctl x uninstall, and to answer your question:
reinstall the gateway(s) for the previous revision manually
means install your gateways the same way you did when deploying Istio for the first time. If you used the default configuration, and did not create any custom gateways, you can just use yaml files that come with the installation files of lower version.
- Create kubernetes secret for docker registry - Terraform
- Kubectl error: memcache.go:265] couldn’t get current server API group list: Get
- Invalid resource manager ID in primary checkpoint record
- How can I view the config details of the current context in kubectl?
- Webhook call fails while deploying a new Couchbase cluster
- How can we create same resource in multiple terraform providers?
- 502 Bad Gateway on Azure Ingress for Kubernetes Service /microservice1/hello (Quarkus Application)
- Can't access a Minikube's NodePort service from outside
- Infinispan dual roles on Kubernetes as a Vert.x cluster manager and cache manager. Cache not being shared problem
- how to use kubectl command with flag --selector?
- Spring Boot with Kubernetes Issue for Prometheus (Cannot file prometheus yml file for configuration)
- EKS cluster pods unable to connect to internet
- kubernetes pod logging broken with journald logging driver
- Helm configMap support for binary files
- Usage of variables in Helm
- How to share a cephfs volume between pods in different k8s namespaces
- is it possible to use Gitlabci parallel matrix to build docker image with certain time delay for each image?
- How to resolve invalid character included yaml attributes from kubernetes
- Why kubectl exec --username=root does not work?
- How to enter a pod as root?
- How to exclude a resource when using namespace in kustomization.yaml?
- Kubernetes Ingress Routing HTTP with Kestrel
- How to pass brackets and other special characters using —set in Helm
- Problem installing Issuer (cert-manager) inside GKE cluster tls: failed to verify certificate: x509: certificate signed by unknown authority
- couldn't get current server API group list: the server has asked for the client to provide credentials error: You must be logged in to the server
- Spring Boot Application Fails to Create/Replace ConfigMap in Minikube After Upgrade
- GKE Autopilot: How to add/manage SSL Certificate to GKE autopilot
- Is it possible to allow egress traffic by hostname?
- helm chart error can't evaluate field Values in type interface {}
- What is the difference between Helm and Kustomize?