This required a lot of bumps to keep things working:
- `siderolabs/tools` to `v1.12.0-1-g188885e` (this is bumped to the HEAD
of the `release-1.12` branch to include Go bump)
- `siderolabs/pkgs` to `v1.12.0`
Additionally, due to
52db66e8d9m,
had to replace `protoc` with `buf` and add an appropriate `buf.yaml`.
Also includes some `talos@v1.12.0-beta.0` related non-trivial changes in
sfyra tests.
Signed-off-by: Laura Brehm <laurabrehm@hey.com>
This commit implements strategic merge patch by reusing Talos
configpatcher patch handling. Furthermore, a unit test scenario has been
added to ensure its functionality.
Fixes: #1259
Signed-off-by: Ksawery Kuczyński <ksawerykuczynskikk@gmail.com>
Signed-off-by: Andrey Smirnov <andrey.smirnov@siderolabs.com>
This is final part of #735, previous part in #823.
This imports all remaining changes with some fixups minus the webhook
changes.
This change adds detailed hardware information to the Server CRD.
Hardware info is extracted by the agent from SMBIOS.
The ServerClass CRD is also updated so more precise qualifiers can be used.
Co-authored-by: Gerard de Leeuw <gdeleeuw@leeuwit.nl>
Signed-off-by: Gerard de Leeuw <gdeleeuw@leeuwit.nl>
Signed-off-by: Andrey Smirnov <andrey.smirnov@talos-systems.com>
This PR fixes the gpg check as well as adds the clusterctl workaround to
make sfyra build properly.
Signed-off-by: Spencer Smith <spencer.smith@talos-systems.com>
Includes bumps to the pkgs/tools for Go 1.17.7 and new kernel version.
Added clusterctl move label to the `siderolink` root secret.
Signed-off-by: Andrey Smirnov <andrey.smirnov@talos-systems.com>
Use same matcher in both log receiever and event receiver. Fix panic in
event receiver.
Bump CAPI and remove cert-manager temporary hack.
Signed-off-by: Andrey Smirnov <andrey.smirnov@talos-systems.com>
If the environment is not ready, don't boot from it, as it would fail
anyways.
Fixes#658
Signed-off-by: Andrey Smirnov <andrey.smirnov@talos-systems.com>
Now `PXEBooted` condition is derived from `TalosInstalled` condition of
the linked `ServerBinding`.
This makes the node to use `pxe` boot until Talos installation succeeds.
Signed-off-by: Artem Chernyshev <artem.chernyshev@talos-systems.com>
- `TalosConfigLoaded` is set to false when the config load has failed.
- `TalosConfigValidated` is set to false when the config validation
fails on the node.
- `TalosInstalled` is set to true/false when talos installer finishes.
All conditions are set by the adapter on the `ServerBinding`, then
copied to the `MetalMachine`.
Signed-off-by: Artem Chernyshev <artem.chernyshev@talos-systems.com>
Get addresses from the events coming into events sink container.
Add v1beta1 support.
Update `metalmachine_controller` to set providerID earlier and add
ProviderSet condition to be set after the cluster nodes actually get
patched with the provider ID as well.
The condition is required to trigger reconcile on CAPI `machines`,
otherwise it never picks `Node` ready condition.
Signed-off-by: Artem Chernyshev <artem.chernyshev@talos-systems.com>
This PR updates the naming of the two apps that make up sidero for
consistency. It also updates things like the Makefile, Dockerfile, and
kustomize to make sure image names are the same as these app names.
Signed-off-by: Spencer Smith <robertspencersmith@gmail.com>
This merges metadata server component into `metal-controller-manager`.
iPXE server, metadata server and gRPC server share a single HTTP
endpoint now.
For the users, it means that Sidero requires only a single port to be
exposed.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This allows to override default options (e.g. `console=ttyS0`) with
provided values. Make sure that full value is given for each option, as
any default values for the option are dropped if specified in
`--extra-agent-kernel-args`, e.g.
`--extra-agent-kernel-args "console=tty0 console=ttyS1 panic=5"`.
See also https://github.com/talos-systems/go-procfs/pull/5Fixes#251
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This injects power management failures in Sfyra (under test). By default
failures are not injected.
Two class of failures are supported:
* explicit (when client returns an error)
* silent (when client doesn't do the job, but error is not returned)
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
In 0.7.0-alpha.4, asset names got changed slightly.
This bumps bootstrap cluster resources as 2GiB is not enough for CAPI.
Fixes a small bug with asset filenames in environment controller.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
