Bug 1838512

Summary: oc doesn't work on Mac OS through VPN
Product: OpenShift Container Platform Reporter: David Caldwell <dcaldwel>
Component: ocAssignee: Mike Dame <mdame>
Status: CLOSED WONTFIX QA Contact: zhou ying <yinzhou>
Severity: medium Docs Contact:
Priority: low    
Version: 4.4CC: aos-bugs, jokerman, maszulik, mdame, mfojtik
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2021-05-17 15:00:41 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
route none

Description David Caldwell 2020-05-21 10:12:08 UTC 
Description of problem:

OpenShift Client for Mac OS cannot contact to clusters behind the corporate VPN. kubectl works fine.


Version-Release number of selected component (if applicable):

OCP 4.4


How reproducible:

Using MacOS behind a VPN, try to use the oc client to connect to and perform oc commands.


Actual results:

Host unavailable error.


Expected results:

Host successfully resolves.


Additional info:

The issue described: https://github.com/kubernetes/kubectl/issues/48
And solution already implemented in kubectl: https://github.com/kubernetes/release/issues/469
I have attached a BlueJeans recording of the customer reproducing the issue.
From the above github links and from what the customer says in the BlueJeans chat, it looks like recompiling oc client with cgo enabled should fix this.
Comment 2 Maciej Szulik 2020-05-21 10:31:04 UTC 
Mike can you check this one out since I remember you being on Mac ;-)
Comment 4 David Caldwell 2020-05-28 07:07:27 UTC 
@Mike

Thanks for your help with this issue. I asked the customer if he could compile the OC client with CGO_ENABLED=1. This works for him as a temporary workaround using client release 4.4. 

Customer's OS is MacOS Catalina 10.15.4

Customer's OC client version is 4.4.4
Comment 5 Mike Dame 2020-05-28 13:52:50 UTC 
How is the customer currently building oc when the do see the issue? Are they compiling it themselves from source, or downloading a binary from a specific place/release for example?

As I tested this with several published binaries, as well as building from source, and wasn't able to reproduce the bug on MacOS Catalina then if there is any other information about their specific oc it could help me try to reproduce it. Would it be possible to even share the exact binary they are running when they see the issue?
Comment 9 David Caldwell 2020-05-29 09:48:20 UTC 
Created attachment 1693315 [details]
route
Comment 17 Mike Dame 2020-06-09 20:54:34 UTC 
I have spoken with the release automation team who informed me that a similar request has already been made (for a different tool) to allow CGO_ENABLED builds for MacOS binaries (see https://issues.redhat.com/browse/ART-1674)

I have commented in that Jira issue that this change would also fix this bug with oc. I will continue to track this Jira issue to make sure that when the fix is available that it is enabled for MacOS oc binaries as well.
Comment 20 Mike Dame 2020-08-20 17:52:37 UTC 
I recently requested an update on the progress of this in the tracking card (https://issues.redhat.com/browse/ART-1674). It seems that the process for approving these build images needs to be re-started for 4.6, trying to find if that will be made available for earlier versions.
Comment 21 Mike Dame 2020-08-21 13:27:56 UTC 
I’m adding UpcomingSprint, because I was occupied by fixing bugs with higher priority/severity, developing new features with higher priority, or developing new features to improve stability at a macro level. I will revisit this bug in a future sprint.
Comment 22 Mike Dame 2020-09-10 16:02:05 UTC 
It looks like cross-compilation builder images were added in https://github.com/openshift/ocp-build-data/pull/627. We will look at what, if anything, is necessary to enable these for delivery in an upcoming sprint
Comment 23 Mike Dame 2020-10-02 21:18:59 UTC 
I’m adding UpcomingSprint, because I was occupied by fixing bugs with higher priority/severity, developing new features with higher priority, or developing new features to improve stability at a macro level. I will revisit this bug in an upcoming sprint.
Comment 24 David Caldwell 2020-10-08 06:38:03 UTC 
Thank you for the update - it is appreciated.
Comment 26 Mike Dame 2020-11-17 20:20:55 UTC 
I've opened a small change to oc which I believe should resolve this problem now that the c cross-compilers are available in build images (https://github.com/openshift/oc/pull/644)
Comment 28 Mike Dame 2020-12-04 17:56:31 UTC 
There's some feedback I received on that PR that this may not be able to merge until we have at least Go 1.15.4 available. Will keep following this
Comment 29 Mike Dame 2021-01-17 02:21:53 UTC 
This is still blocked on limits in our build process, discussion will continue in the linked PR
Comment 30 Mike Dame 2021-03-19 19:17:12 UTC 
Based on the requirements for this as described in https://github.com/openshift/oc/pull/644#issuecomment-790093212, I don't think we are considering pursuing fixing this at this time.
Comment 31 Mike Dame 2021-05-17 15:00:41 UTC 
As I mentioned in the above comment, the requirements to fix this are outside the scope of what we are reasonably able to accomplish. Since there has been no updates or feedback since that comment, I am going to close this bugzilla.

Feel free to reopen if it is determined that we should continue pursuing this.