Skip to content

Conversation

@cfBrianMiller
Copy link
Member

No description provided.

wking and others added 30 commits January 21, 2020 13:51
The plural form was deprecated in 2e7ecd3 (types: rename network
type to match Network.config.openshift.io, 2019-05-04, #1356).
F-strings are new in 3.6 [1], and my old RHEL 7.5 CSB has the old
Python 3.4.9 by default.  I'm probably just way behind the times, but
it doesn't cost much to use the older .format() to make this
compatible with all of Python 3 ;).

[1]: https://www.python.org/dev/peps/pep-0498/
This opts us in to some of the features from
https://cloud.google.com/security/shielded-cloud/shielded-vm
Specifically with this, we get a vTPM device.

And what's nice about having a TPM device is that we can start
to optionally make use of TPM devices in OpenShift which
will then work on both bare metal *and* in GCP.

Closes: #2546
Add NoAllowedAddressPairs parameter.
Bug 1814172: [ovirt] Cannot run openshift-install as a non-root user
gcp: Flag RHCOS with SECURE_BOOT and UEFI_COMPATIBLE
This commit fix an indentation on the UPI destroy playbook
for loadbalancers.
Add link to the github issue tracking re-enabling of glance image
import:

#3403
This commit fixes the problem where S3 bucket deletion would fail if it had
versioned objects. deleteS3() now does a mass delete of any versioned objects
in the S3 bucket before deleting the bucket.
OWNERS: include component for image tracking by ART
This commit adds validations that all provided network and security
group ids are UUID v4 strings.
Bug 1810840 : Quote user and password values for good UX and avoid co…
bug 1815518: pkg/destroy/aws: delete versioned objects from s3 buckets
This logic became a method in 1ab1cd3 (types: add ClusterDomain
helper for InstallConfig, 2019-01-31, #1169); so we can drop the
validation-specific helper which was from bf3ee03 (types: validate
cluster name in InstallConfig, 2019-02-14, #1255).  Or maybe we never
needed the validation-specific helper ;).
OpenStack: Link to GH issue for glance image import
When running an Azure uninstall for a cluster that has already been
deleted, the uninstall fails due to an error deleting the resource
group. This change accepts an already-deleted resource group as a
successful uninstall of the resource group.

https://issues.redhat.com/browse/CO-863
On ppc64le there were OOM kills being observed during the bootstrap process
because of insufficient memory and bumping the memory seemed to solve the problem.
The libvirt defaults for the master and worker memory are 7G and 5G respectively,
so setting the boostrap default to 5G for ppc64le. ppc64le uses 64K pages rather than
the default 4K page size and thus requires more memory.
Having playbooks names with numbers can be misleading,
as the the down playbooks are not triggered on the order
indicated on the playbook name.
New validation rule to verify that the hosts specified in the baremetal platform configuration is at least the amount of configured replicas for ControlPlane and Compute nodes
This commit ensures the lb ID is used instead of the name
when deleting a lb, as openstack allows multiple load balancers
with same name to coexist. Also, it ensures the deletion of lb
with status different than ERROR or ACTIVE to be skipped.
Adds a new feature which prints a summary of the time it took in
each stage of the installer after the stage has been completed.

Summary is currently supported for the create, wait-for and the
delete command. The summary is printed at the completion of the
command and will show the time elapsed for each of the stages
of the execution.

Create command has the following stages:-
. Infrastructure
. API
. Bootstrap Complete
. Destroy bootstrap complete.
. Cluster Operator Complete.

Destroy command has the following stages:-
. Cluster
. Bootstrap

The wait-for command has the following stages:-
- Bootstrap-complete command
	. Bootstrap
	. API
- Install-complete command
	. Cluster
	. Console
Fix documentation to no longer reflect needless requirement of cloning
the installer repository to a directory in $GOPATH/src.

[issue #3422]

* Tags: kind/documentation platform/none size/XS
LorbusChris and others added 29 commits April 19, 2020 12:23
Add instructions for UPI install on Power
Dockerfile.upi.ci: Add GCP UPI required binaries.
Bug 1819492: vsphere upi and metal: terraform 0.12.x update, general updates and reorg
When gathering bootstrap logs, it's not clear where the log bundle is written.
This prints out the absolute path to the log bundle.
No functional changes.

An Ansible step mentions Server groups, while it gathers a token for
calls unrelated to server groups.
Bug 1826739: openstack UPI: Fix step name
Fixes the error message that occurs after the region selection when an
invalid project ID is entered. Solved by picking up the list of projects
that the user has in their GCP account and offers a select option on the
project entry to select the one that they want. Reduces the chance of
picking an invalid project ID and avoids the error.

CloudResourceManager is used to pick up the list of projects for current
user and survey is populated with the values obtained.

Generated the gcpclient_generated.go mock file using mockgen and replaced
the existing file to add the GetListOfProjects function. Running
hack/go-genmock.sh did not work after I changed the interface and ran it.
bug 1821932: cmd/openshift-install/gather: clarify location of log bundle
For the provisioning config, it's allowable for provisioningDHCPRange to
be empty. If it is, the current template produces this YAML:

```yaml
spec:
   provisioningDHCPRange:
```

Unfortunately, in YAML-land the above is null not empty string, so when
you apply it, you get complaints:

```
Provisioning.metal3.io "provisioning-configuration" is invalid:
spec.provisioningDHCPRange: Invalid value: "null":
spec.provisioningDHCPRange in body must be of type string: "null"
```

This change ensures all string values are quoted, which will result in a
correct value for provisioningDHCPRange when empty.
Bug 1826414: Added validation for GCP Project ID access
Documentation change.

Add a comment explaining why we are installing the AWS command-line
client in the CI image for OpenStack.
The installer creates a server group as a side effect of the manifest
generation.

Prior to this patch, the UPI playbooks created yet another server group
and left the installer-created one unused.

With this patch, the UPI playbooks reuse the server group created by the
installer, rather than creating an additional one.
In the case of multiple added networks the tag
<infraID>-primaryClusterNetwork should allow
cluster-api-provider-openstack to define which
IP address to set as the primary one for machines.

Now CAPO can't do this, because Neutron returns the
list of networks in alphabetical order.
Bug 1827070: openstack: Explain why awscli in the CI image
Bug 1824426: tag primary OpenStack networks
Bug 1823359: baremetal: update provisioning CR to quote strings
Bug 1827067: openstack UPI: Server group name to match IPI
Bug 1826895: rhcos: bump RHCOS boot image to 44.81.202004250133-0
Bug 1794839: Azure masters should correctly support service load balancers
Using jq will simplify the CI scripts where we currently use `awk` to
parse JSON.
For a successful installation, it is required to be able to tag network
resources.
Bug 1827667: openstack: Add jq to the CI image
Bug 1826317: openstack docs: Add tagging as a requirement
Bug 1829409: Update vsphere machine api to beta
@cfBrianMiller cfBrianMiller merged commit db8614b into CloudFitSoftware:master Apr 30, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.