Skip to content

CHANGELOG: Document changes since v0.3.0 #595

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.

Sign up for GitHub

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

Merged
openshift-merge-robot merged 1 commit into from
Nov 22, 2018
Merged

CHANGELOG: Document changes since v0.3.0 #595

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
162 changes: 158 additions & 4 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,151 @@ All notable changes to this project will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).

## 0.4.0 - 2018-11-22

### Added

- The creation targets have been moved below a new `create` subcommand
(e.g. `openshift-install create cluster` instead of the old
`openshift-install cluster`). This makes them easier to distinguish
from other `openshift-install` subcommands and also mirrors the
approach taken by `destroy` in 0.3.0.
- A new `manifest-templates` target has been added to `create`,
allowing users to edit templates and have descendant assets
generated from their altered templates during [a staged
install](docs/user/overview.md#multiple-invocations).
- [The ingress operator][ingress-operator] is no longer masked. The
old Tectonic ingress operator has been removed.
- The [the registry operator][registry-operator] has been added, and
the kube-addon operator which used to provide a registry (among
other things) has been removed.
- The [checkpointer operator][checkpointer-operator] is no longer
masked. It runs on the production cluster, but not on the bootstrap
node.
- Cloud credentials are now pushed into a secret where they can be
consumed by cluster-API operators and other tools.
- OpenStack now has `destroy` support.
- We log verbosely to `${INSTALL_DIR}/.openshift_install.log` for most
operations, giving access to the logs for troubleshooting even if
you neglected to run with `--log-level=debug`.
- We've grown [troubleshooting
documentation](docs/user/troubleshooting.md).

### Changed

- For consistency with 0.3.0's `destroy`, the the old
`install-config`, `manifests`, `ignition-configs`, and `cluster`
commands have been pushed down under a new `create` subcommand. For
example, you should now use `openshift-install create cluster`
instead of `openshift-install cluster`. The old commands are
deprecated.
- The `create cluster` subcommand now waits for the
`bootstrap-complete` event and automatically removes the bootstrap
assets after receiving it. This means that after `create cluster`
returns successfully, the cluster has its production control plane
and topology (although there may still be operators working through
their initialization). The `bootstrap-complete` event was new in
0.3.0, and it is now pushed at the appropriate time (it was too
early in 0.3.0). The `destroy bootstrap` subcommand is still
available, to allow users to manually trigger bootstrap deletion if
the automatic removal fails for whatever reason.
- On AWS, bootstrap deletion now also removes the S3 bucket used for
the bootstrap node's Igntition configuration.
- Asset state is preserved even while moving backwards through [a
staged install](docs/user/overview.md#multiple-invocations). For
example:

```sh
openshift-install --dir=example create ignition-configs
openshift-install --dir=example create install-config
```

now preserves the full state including the generated Ignition
configuration. In 0.3.0, the `install-config` call would have
removed the Ignition configuration and other downstream assets
from the stored state.
- Some asset state is removed by successful `destroy cluster` runs.
This reduces the change of contaminating future cluster creation
with assets left over from a previous cluster, but users are [still
encouraged](README.md#cleanup) to remove state between clusters to
avoid accidentally contaminating the subsequent cluster's state.
- etcd discovery now happens via `SRV` records. On libvirt, this
requires a new Terraform provider, so users with older providers
should [install a newer
version](docs/dev/libvirt-howto.md#install-the-terraform-provider).
This also allows all masters to use a single Ignition file.
- On AWS, the API and service load balancers have been changed from
[classic load balancers][aws-elb] to [network load
balancers][aws-nlb]. This should avoid [some latency issues we were
seeing with classic load balancers][aws-elb-latency], and network
load balancers are cheaper.
- On AWS, master `Machine` entries now include load balancer
references, ensuring that new masters created by [the AWS
cluster-API provider][cluster-api-provider-aws] will be attached to
the load balancers.
- On AWS and OpenStack, the default network CIDRs have changed to
172.30.0.0/16 for services and 10.128.0.0/14 for the cluster, to be
consistent with previous versions of OpenStack.
- The bootstrap kubelet is no longer part of the production cluster.
This reduces complexity and keeps production pods off of the
temporary bootstrap node.
- [The cluster-version operator][cluster-version-operator] now runs in
a static pod on the bootstrap node until the production control
plane comes up. This breaks a cyclic dependency between the
production API server and operators.
- The bootstrap control plane now waits for some core pods to come up
before exiting.
- [The machine-API operator][machine-api-operator] now reads the
install-config from the `cluster-config-v1` config-map, instead of
from an operator-specific configuration.
- AWS AMIs and libvirt images are now pulled from the new [RHCOS
pipeline][rhcos-pipeline].
- Updated the security contact information for CoreOS -> Red Hat.
- We push a `ClusterVersion` custom resource. The old `CVOConfig` is
still being pushed, but it is deprecated.
- OpenStack credentials are loaded from standard system paths.
- On AWS and OpenStack, ports 9000-9999 are now open for host network
services.
- Lots of doc and internal cleanup and minor fixes.

### Fixed

- On AWS, `destroy cluster` is now more robust, removing resources with
either the `tectonicClusterID` or `kubernetes.io/cluster/<name>:
owned` tags. It also removes pending instances as well (it used to
only remove running instances).
- On libvirt, `destroy cluster` is now more precise, only removing
resources which are prefixed by the cluster name.
- Bootstrap Ignition edits (via `create ignition-configs`) no longer
suffer from a `worker.ign` dependency cycle, which had been
clobbering manual `bootstrap.ign` changes.
- The state-purging implementation respects `--dir`, avoiding `remove
...: no such file or directory` errors during [staged
installs](docs/user/overview.md#multiple-invocations).
- Cross-filesystem Terraform state recovery during `destroy bootstrap`
no longer raises `invalid cross-device link`.
- Bootstrap binaries are now located under `/usr/local/bin`, avoiding
SELinux violations on RHEL 8.

### Removed

- All the old Tectonic operators and the `tectonic-system` namespace
have been removed.
- On libvirt, the image URI prompt has been removed. You can still
control this via the `OPENSHIFT_INSTALL_LIBVIRT_IMAGE` environment
variable, but too many users were breaking their cluster by pointing
the installer at an outdated RHCOS, so we removed the prompt to make
that knob less obvious.
- On libvirt, we've removed `.gz` suffix handling for images. The new
RHCOS pipeline supports `Content-Encoding: gzip`, so the
suffix-based hack is no longer necessary.
- The `destroy-cluster` command, which was deprecated in favor of
`destroy cluster` in 0.3.0, has been removed.
- The creation target subcommands of `openshift-install` have been
removed. Use the target subcommands of `create` instead
(e.g. `openshift-install create cluster` instead of
`openshift-install cluster`).

## 0.3.0 - 2018-10-22

### Added
Expand All @@ -22,8 +167,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).

- The installer creates worker `MachineSet`s, instead of leaving that to
[the machine-API operator][machine-api-operator].
- Creates master `Machine`s and tags masters to be picked up by the
[AWS cluster-API provider][cluster-api-provider-aws].
- The installer creates master `Machine`s and tags masters to be
picked up by the [AWS cluster-API
provider][cluster-api-provider-aws].

### Changed

Expand Down Expand Up @@ -63,8 +209,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
and the new service is labeled so [Prometheus][] will scrape it.
- The `service-serving-cert-signer-signing-key` secret is now
available in the `openshift-service-cert-signer` namespace, which
gives [the service-serving cert signer][] the keys it needs to mint
and manage certificates for Kubernetes services.
gives [the service-serving cert signer][service-serving-cert-signer]
the keys it needs to mint and manage certificates for Kubernetes
services.
- The etcd-serving certificate is now passed through to [the
kube-controller-manager operator][kube-controller-manager-operator].
- We disable some components which [the cluster-version
Expand Down Expand Up @@ -136,12 +283,19 @@ installer and follow along as it guides you through the process.
The `tectonic` command and tarball distribution are gone. Please use
the new `openshift-install` command instead.

[aws-elb]: https://docs.aws.amazon.com/elasticloadbalancing/latest/classic/introduction.html
[aws-elb-latency]: https://github.com/openshift/installer/pull/594#issue-227786691
[aws-nlb]: https://docs.aws.amazon.com/elasticloadbalancing/latest/network/introduction.html
[checkpointer-operator]: https://github.com/openshift/pod-checkpointer-operator
[cluster-api-provider-aws]: https://github.com/openshift/cluster-api-provider-aws
[cluster-version-operator]: https://github.com/openshift/cluster-version-operator
[dot]: https://www.graphviz.org/doc/info/lang.html
[ingress-operator]: https://github.com/openshift/cluster-ingress-operator
[kube-apiserver-operator]: https://github.com/openshift/cluster-kube-apiserver-operator
[kube-controller-manager-operator]: https://github.com/openshift/cluster-kube-controller-manager-operator
[machine-api-operator]: https://github.com/openshift/machine-api-operator
[machine-config-operator]: https://github.com/openshift/machine-config-operator
[Prometheus]: https://github.com/prometheus/prometheus
[registry-operator]: https://github.com/openshift/cluster-image-registry-operator
[rhcos-pipeline]: https://releases-rhcos.svc.ci.openshift.org/storage/releases/maipo/builds.json
[service-serving-cert-signer]: https://github.com/openshift/service-serving-cert-signer