"Failed to get workspace configuration" message when opening workspace on OpenShift behind proxy #14676
Comments
tolusha
added
kind/bug
|
@andy316x Thank you for your report. OpenShift templates in che repo are not the recommended way to install che. Can you try to follow official documentation https://www.eclipse.org/che/docs/che-7/installing-che-on-openshift-3-using-the-operator/? |
skabashnyuk
added
kind/question
|
@skabashnyuk thanks for your response, still waiting on checking the issue as I needed to upgrade OpenShift to 3.11 in order to make use of chectl for deployment. Nearly there I think, OpenShift is upgraded but chectl doesn't seem to want to work, not sure if you have seen this issue before, I am getting:
when attempting to run I will get back to you with a response as soon as I have been able ot install using the operator. |
|
@skabashnyuk I now have Che deploying using chectl on OpenShift 3.11, but I do not see an option to set proxy settings in chectl in the same way that you were able to with the deployment templates, is setting proxy details not supported in chectl? |
|
@andy316x you can try configure proxy settings in CheCluster cr, see https://github.com/eclipse/che-operator/blob/master/deploy/crds/org_v1_che_cr.yaml#L40-L48 |
|
@skabashnyuk I have now successfully deployed Che using chectl - thanks for your support on the additional configuration, unfortunately I still get the exact same error as before, i.e. "Failed to get workspace configuration" once the IDE has loaded - making the workspace useless. I am convinced that this is down to the version of Axios that is being used in the che-theia container, the version being used in 0.18.0, which respects http_proxy and https_proxy, however it does not implement no_proxy, this was introduced in 0.19.0. This would line up with the behaviour I am seeing, it is hitting my corporate proxy when attempting to read the workspace configuration from: http://che-che.apps.company.aws.local/api/workspace/[workspaceId] Even though I have che-che.apps.company.aws.local in my no_proxy environment variable in the container. no_proxy was added to Axios in 0.19.0: I think upgrading to the latest version of Axios in the che-workspace-client component used by che-theia will resolve this issue. |
New axios version provides support for no_proxy variable and solves the problem outlined at eclipse-che/che#14676
New axios version provides support for no_proxy variable and solves the problem outlined at eclipse-che/che#14676 Signed-off-by: Diego Dompe <ddompe@gmail.com>
|
Quick Update: The upgrade to axios is shipping since 7.5.0, and seems to work except that the no_proxy doesn't support wildcards like '*.domain.com', so I had to specify the full exception for the api server. After that I'm still having troubles because theia is not trusting my root-ca, so I get 'unable to verify the first certificate' on the logs. Trying to figure out how to make it ignore it or trust it... |
|
@ddompe great to hear that this has gone in, I will attempt to give that a go. I was able to get a workaround for this by putting everything through my own proxy, I then applied the no proxy rules there where I wasn’t dependent on axios, but this is a kludge I would like to get rid of. With respect to your wildcards not working, when investigating this issue initially I remember the no proxy feature axios implemented had a limitation where it would only apply no proxy rules for wildcards one level deep, so you could still be in luck. Check this open issue for more details: On your CA trusting issue, I experience the same problem, which is an issue for us with our private CA as we currently can’t run with SSL on Openshift, so I would be interested how you get on with this. Do you know if there is an open ticket? If not, I will create one.. |
|
Issues go stale after Mark the issue as fresh with If this issue is safe to close now please do so. Moderators: Add |
che-bot
added
the
lifecycle/stale
Summary
"Failed to get workspace configuration" message when opening workspace on OpenShift behind proxy when loading any of the Che stacks, causing the workspace not to be able to load consoles etc.
Installation
On AWS EC2 deployed the following 1 master, 1 infra, 2 worker node OpenShift cluster (with self signed certs)
OpenShift: v3.9.0+0fd118e-40
Kubernetes: v1.9.1+a0ce1bc657
Using the latest OpenShift deployment templates in the Che Git repo, deploy a multi-user, SSL version of the nightly build of Che, http_proxy set, no_proxy set to .apps.company.aws.local,.compute.internal
Steps to reproduce
Kubernetes events:
Che server events:
(All before starting the workspace)
Che theia workspace container logs:
Note that it gets past this error if I remove the proxy configuration from the che-theia container
I have also provided comments in linked issue #14052. After some digging looks like a potential issue could be the version of axios being used (v0.18.0) as it seems not to support the no_proxy parameter.
The text was updated successfully, but these errors were encountered: