Bug 1564765
Summary: | Configuring for vSphere cloud provider fails to add nodes to cluster | ||
---|---|---|---|
Product: | OpenShift Container Platform | Reporter: | Timothy Ling <tling> |
Component: | Documentation | Assignee: | Vikram Goyal <vigoyal> |
Status: | CLOSED DEFERRED | QA Contact: | Vikram Goyal <vigoyal> |
Severity: | unspecified | Docs Contact: | Vikram Goyal <vigoyal> |
Priority: | unspecified | ||
Version: | 3.7.0 | CC: | aos-bugs, aos-storage-staff, bchilds, dphillip, jokerman, mmccomas, wmeng, yves.vogl |
Target Milestone: | --- | ||
Target Release: | --- | ||
Hardware: | x86_64 | ||
OS: | Linux | ||
Whiteboard: | |||
Fixed In Version: | Doc Type: | If docs needed, set a value | |
Doc Text: | Story Points: | --- | |
Clone Of: | Environment: | ||
Last Closed: | 2019-11-21 12:49:15 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: |
Description
Timothy Ling
2018-04-07 14:52:21 UTC
I don't have vmWare at hand to check this, but all other cloud providers *require* the machine hostname to be equal to name of the VM in vmWare. Did this recommendation work: https://docs.openshift.com/container-platform/3.7/install_config/configuring_vsphere.html#vsphere-applying-configuration-changes ? If not, then we need to update it to something that works. Or we might update the docs to recommend installation OpenShift on vmWare machines already with the right hostnames. You can fix the hostnames later with some manual effort and it seems to me you did it correctly. It's not a workaround, it's the fix. @Davis, do you have any other insights who/how should set hostnames when installing OpenShift on vmWare? Is our documentation accurate? @Jan, yes that is correct. Upon implementing the cloud provider configuration for vSphere the hostname in OpenShift is changes to the virtual machine name. We circumvent the problem by using hostvars to assign openshift_name and set it to the inventory_hostname in ansible. Here is an example with a dynamic inventory role in Ansible: https://github.com/openshift/openshift-ansible-contrib/blob/master/reference-architecture/vmware-ansible/playbooks/roles/instance-groups/tasks/main.yaml#L35-L42 Here is an example with a static inventory in Ansible: https://gist.github.com/dav1x/b374f2e2becbeae1b6efd6c36aefd1ce The cluster installation and configuration just needs to align with what the cloud provider expects. We could probably clarify that in the docs. I'll make myself a task to do so. I am re-assigning the bug to Docs then. BTW, should we update docs of all the other cloud providers? It's exactly the same on all of them, VM name should be the same as hostname. Thanks everyone for the input. I can confirm that when we renamed the VM names to their matching FQDNs and reinstalled OCP, everything worked as expected. If this requirement is documented in the other cloud providers, I must have missed it and going forward I think it would be very helpful to have this called out as a requirement in the documentation for using a vSphere cloud provider. Hi, I'd just like to report exactly the same issue. While not having tried the provided workaround I'd like to offer help if needed. We've created a fully reproducable setup with Packer, Terraform and Ansible. Kind regards, Yves As specified above, this is more of a documentation issue. As of Openshift 3.11, the VM name requirement is now documented, but it is a little hidden. Check https://docs.openshift.com/container-platform/3.11/install_config/configuring_vsphere.html#install-config-configuring-vsphere Then scroll down to "Configuring OpenShift Container Platform to use vSphere storage" which is about 2/3 the way down the page and it will list the prerequisites. Technically this problem has now been documented, but it might be more helpful if this prerequisite is listed at the start of the page before the actual steps for configuration changes take place. We've found the following issue: - VMware vCenter VMs have the FQDN as name (e.g. node-001.example.com) - Ansible Inventory uses the same FQDN - VMs itself act correctly by just issuing the domain part of the hostname when hostname if specified with "-f" flag I think this could be an issue as when setting the hostname inside the VM to a FQDN (and therefore make no difference between running hostname command with "-f" flag) everything works. OCP 3.6-3.10 is no longer on full support [1]. Marking un-triaged bugs CLOSED DEFERRED. If you have a customer case with a support exception or have reproduced on 3.11+, please reopen and include those details. When reopening, please set the Version to the appropriate version where reproduced. [1]: https://access.redhat.com/support/policy/updates/openshift |