Closed andreclaudino closed 4 years ago
Ah yes... It's likely that inverse proxy does not support South America, yet.
Is it a necessity to deploy clusters over there for you? If so we'll file a FR to inverse proxy team.
@andreclaudino Below is the currently supported regions. @rmgogogo I think we need to document that in the website as well.
us-central1
us-central2
us-west1
us-east1
us-east4
northamerica-northeast1
europe-west1
europe-west2
europe-west6
asia-southeast1
asia-northeast1
asia-east1
australia-southeast1
+1 for document it in website. @joeliedtke
I will also send out an email to proxy team on whether flexible to add more region.
Ah yes... It's likely that inverse proxy does not support South America, yet.
Is it a necessity to deploy clusters over there for you? If so we'll file a FR to inverse proxy team.
Thanks for answering. Yes, we need to use this in South America. We are using Kubeflow in a production cluster with many other applications from our company, it's impossible for us to change the cluster region only to support kubeflow.
What do we needed to support other regions? I look at the file, and there is an standard in filenames, maybe we can just create these entries automatically for all gcp regions.
What do we needed to support other regions
Unfortunately I don't think there is an easy way for our customers to make changes to support more regions. Inverse proxy is a managed service hosted as a Google managed service, the mapping in https://storage.googleapis.com/ml-pipeline/proxy-agent-config.json
is the list of currently supported GCE regions.
We've filed a FR to inverse proxy team to expand their supported regions. Meanwhile, to use KFP on an arbitrary region, I think one workaround is to deploy a full-fledged Kubeflow, see this guide. KFP is one component of it.
@rmgogogo do we have any other option?
What do we needed to support other regions
Unfortunately I don't think there is an easy way for our customers to make changes to support more regions. Inverse proxy is a managed service hosted as a Google managed service, the mapping in
https://storage.googleapis.com/ml-pipeline/proxy-agent-config.json
is the list of currently supported GCE regions.We've filed a FR to inverse proxy team to expand their supported regions. Meanwhile, to use KFP on an arbitrary region, I think one workaround is to deploy a full-fledged Kubeflow, see this guide. KFP is one component of it.
@rmgogogo do we have any other option?
@numerology, the problem persists if deploying entire kubeflow. In fact, I had many other problems installing kubeflow, then decide use only kfp.
No cheap option for now. Reply from proxy team is that support a new region requires 2-4 weeks (didn't commit to add southameraca-east1 yet but we can help request if no other options).
To get more background, any reason why must use southameraca-east1? is it for some legit consideration?
@rmgogogo I think @andreclaudino answered a similar question above
Thanks for answering. Yes, we need to use this in South America. We are using Kubeflow in a production cluster with many other applications from our company, it's impossible for us to change the cluster region only to support kubeflow.
No cheap option for now. Reply from proxy team is that support a new region requires 2-4 weeks (didn't commit to add southameraca-east1 yet but we can help request if no other options).
To get more background, any reason why must use southameraca-east1? is it for some legit consideration?
As @numerology already said, I have answered your question:
We are using Kubeflow in a production cluster with many other applications from our company, it's impossible for us to change the cluster region only to support kubeflow.
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.
This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.
What steps did you take:
I deployed kubeflow-pipelines standalone in private gke following documentation.
What happened:
The pod proxy-agents keeps
CrashLoopBackOff
, getting logs, I can see the messageValueError: No valid URLs found for zone: southamerica-east1-c
What did you expect to happen:
The pod should run normaly
Environment:
How did you deploy Kubeflow Pipelines (KFP)? I deployed version 0.5.1 following instructions in
https://www.kubeflow.org/docs/pipelines/installation/standalone-deployment/
KFP version: 0.5.1 (9c16e12)
KFP SDK version:
Anything else you would like to add:
This is the complete log:
In fact, the file
https://storage.googleapis.com/ml-pipeline/proxy-agent-config.json
doesn't have an entry forsouthamerica-east1-c
, this looks the problem./kind bug // /area frontend /area backend // /area sdk // /area testing // /area engprod -->