-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Kubeflow Dex Distribution] KF Pipelines 100% Unusable - MULTIPLE PEOPLE REPORTING #5223
Comments
Note most users of Kubeflow experiencing this bug cannot afford to wait 6-9 months for the next release which may or may not address this problem. This bug breaks this product. We need a solution immediately. This part of the product is considered STABLE. |
Kubernetes installed via Kubespray |
Hi @ReggieCarey, sorry for your bad experience. KF 1.2.0 with Dex on K8s is a community maintained distribution. /assign @yanniszark |
@ReggieCarey I'm sorry for your bad experience with 1.2. From Arrikto's side, we supported this distribution until Kubeflow 1.1. In Kubeflow 1.2, we started transitioning out of Our current efforts are focused on releasing 1.3. In Kubeflow 1.3, we plan to support a similar distribution but without |
@yanniszark thanks for looking into this. Some more info. I disabled sidecar injection in the kubeflow namespace. This allowed ml pipeline to connect with MySQL and populate 1 pipeline. I can't get to the pipeline dashboard from the KF dashboard but I can get to it via kubectl proxy. Experiments still don't work. Need to check the cache server status. Q: Why are you consuming MySQL instead of a SQL service? |
So I guess there were some miscommunication, I thought Arrikto was still supporting Kubeflow 1.2 with dex. If that's not the case, we should have deleted the dex distribution from kubeflow.org documentation during the release. |
@Bobgy Just dropping in here quickly as I came across the issue. I think what @yanniszark is said is just that deployment with |
You are right, I was only referring to the distribution |
@Bobgy is there a solution for that? |
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
We are experiencing the same problem in 1.4.0 (on-prem installation with Dex). |
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
For everyone watching, any time you see an error like The most likely case is that there is some asymmetric routing going on. That is, your Pods might be able to create a connection to MySQL, but MySQL might not have a route back to your Pod. Note, MySQL connections are complex and will sometimes initiate new TCP connections back to the client, which will fail in the previous case. I have a more detailed write up on this issue: #3763 (comment) |
Closing this issue as there is no users reporting this since 2022. Feel free to reopen if this issue remains in latest releases. /close |
@rimolive: Closing this issue. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
What steps did you take:
KFP in KF 1.2.0 with Dex on K8s 1.18.9 does not work. I receive an error in the KF dashboard when attempting to view pipelines:
Error: failed to retrieve list of pipelines. Click Details for more information. -> An error occured, no healthy upstream
What happened:
Installed Kubeflow 1.2.0 on-prem as per installation instructions. Any attempt to see pipelines or use pipelines fails.
What did you expect to happen:
I expected to be able to use Pipelines
Environment:
Kubernetes version 1.18.9
Kubeflow version 1.2.0
Installed with Dex, configured after deploy to use LDAP.
ml-pipelines pod fails to start completely. Logs indicate
How did you deploy Kubeflow Pipelines (KFP)?
Installed Kubeflow Pipelines as part of Kubeflow installation for on-prem with dex.
KFP version: 1.0.4
KFP SDK version:
I HAVEN'T GOTTEN FAR ENOUGH TO USE THIS!
Anything else you would like to add:
ml-pipeline pod refuses to run:
Cache Server also is unable to connect to MYSQL
Attempted suggestions for repair (ALL fail - please do not suggest)
The product as advertised online does not work on a vanilla on-prem, K8s installation. It appears to work on GCP, Azure, AwS, and possibly IBM.
Provided diagnostic tools are not compatible with an on-prem installation:
/kind bug
The text was updated successfully, but these errors were encountered: