`ServiceProfilesAccess()`, called by control plane components at
startup, would fail if it encountered an `ErrGroupDiscoveryFailed` from
a GroupVersion request. This error is mostly innocuous, as it returns an
error if any GroupVersion fails. `ServiceProfilesAccess()` only needs to
validate ServiceProfiles are available.
Modify `ServiceProfilesAccess()` to specifically request the
ServiceProfile GroupVersion. Also add Discovery object
(`APIResourceList`) support to `NewFakeClientSets`.
Fixes#2780
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Private k8s clusters, such as the private GKE clusters offered by Google
Cloud, cannot be reached through the current API proxy method.
This commit uses the port forwarding feature already developed.
Also modify dashboard command to not fall back to ephemeral port.
Signed-off-by: Jack Price <jackprice@outlook.com>
CustomResourceDefinition parsing and retrieval is not available via
client-go's `kubernetes.Interface`, but rather via a separate
`k8s.io/apiextensions-apiserver` package.
Introduce support for CustomResourceDefintion object parsing and
retrieval. This change facilitates retrieval of CRDs from the k8s API
server, and also provides CRD resources as mock objects.
Also introduce a `NewFakeAPI` constructor, deprecating
`NewFakeClientSets`. Callers need no longer be concerned with discreet
clientsets (for k8s resources vs. CRDs vs. (eventually)
ServiceProfiles), and can instead use the unified `KubernetesAPI`.
Part of #2337, in service to multi-stage check.
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Numerous codepaths have emerged that create k8s configs, k8s clients,
and make k8s api requests.
This branch consolidates k8s client creation and APIs. The primary
change migrates most codepaths to call `k8s.NewAPI` to instantiate a
`KubernetesAPI` struct from `pkg`. `KubernetesAPI` implements the
`kubernetes.Interface` (clientset) interface, and also persists a
`client-go` `rest.Config`.
Specific list of changes:
- removes manual GET requests from `k8s.KubernetesAPI`, in favor of
clientsets
- replaces most calls to `k8s.GetConfig`+`kubernetes.NewForConfig` with
a single `k8s.NewAPI`
- introduces a `timeout` param to `k8s.NewAPI`, currently only used by
healthchecks
- removes `NewClientSet` in `controller/k8s/clientset.go` in favor of
`k8s.NewAPI`
- removes `httpClient` and `clientset` from `HealthChecker`, use
`KubernetesAPI` instead
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Add config.linkerd.io/disable-identity annotation
First part of #2540
We'll tackle support for `--disable-identity` in `linkerd install` in a
separate commit.
Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
Fixes#2465
* Add check for unschedulable pods and psp issues (#2465)
* Return error reason and message on pod or node failure
Signed-off-by: Gaurav Kumar <gaurav.kumar9825@gmail.com>
The `linkerd upgrade` command read the control-plane's config from
Kubernetes, which required the environment to be configured to connect
to the appropriate k8s cluster.
Intrdouce a `linkerd upgrade --from-manifests` flag, allowing the user
to feed the output of `linkerd install` into the upgrade command.
Fixes#2629
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Add validation webhook for service profiles
Fixes#2075
Todo in a follow-up PRs: remove the SP check from the CLI check.
Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
* Define proxy version override annotation
* Don't override global linkerd version during inject
This ensures consistent usages of the config.linkerd.io/linkerd-version and
linkerd.io/proxy-version annotations. The former will only be used to track
overridden version, while the latter shows the cluster's current default
version.
* Rename proxy version config override annotation
Signed-off-by: Ivan Sim <ivan@buoyant.io>
* Disable external profiles by default
* Rename the --disable-external-profiles flag to --enable-external-profiles
Signed-off-by: Ivan Sim <ivan@buoyant.io>
The `install` command errors when the deploy target contains an existing
Linkerd deployment. The `upgrade` command is introduced to reinstall or
reconfigure the Linkerd control plane.
Upgrade works as follows:
1. The controller config is fetched from the Kubernetes API. The Public
API is not used, because we need to be able to reinstall the control
plane when the Public API is not available; and we are not concerned
about RBAC restrictions preventing the installer from reading the
config (as we are for inject).
2. The install configuration is read, particularly the flags used during
the last install/upgrade. If these flags were not set again during the
upgrade, the previous values are used as if they were passed this time.
The configuration is updated from the combination of these values,
including the install configuration itself.
Note that some flags, including the linkerd-version, are omitted
since they are stored elsewhere in the configurations and don't make
sense to track as overrides..
3. The issuer secrets are read from the Kubernetes API so that they can
be re-used. There is currently no way to reconfigure issuer
certificates. We will need to create _another_ workflow for
updating these credentials.
4. The install rendering is invoked with values and config fetched from
the cluster, synthesized with the new configuration.
Have the Webhook react to pod creation/update only
This was already working almost out-of-the-box, just had to:
- Change the webhook config so it watches pods instead of deployments
- Grant some extra ClusterRole permissions
- Add the piece that figures what's the OwnerReference and add the label
for it
- Manually inject service account mount paths
- Readd volumes tests
Fixes#2342 and #1751
Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
This change reintroduces identity hinting to the destination service.
The Get endpoint includes identities for pods that are injected with an
identity-mode of "default" and have the same linkerd control plane.
A `serviceaccount` label is now also added to destination response
metadata so that it's accessible in prometheus and tap.
This change adds a new `linkerd2-proxy-identity` binary to the `proxy`
container image as well as a `linkerd2-proxy-run` entrypoint script.
The inject process now sets environment variables on pods to support
identity, including identity names for the destination and identity
services.
As the proxy starts, the identity helper creates a key and CSR in a
tmpfs. As the proxy starts, it reads these files, as well as a
serviceaccount token, and provisions a certificate from controller.
The proxy's /ready endpoint will not succeed until a certificate has
been provisioned.
The proxy will not participate in identity with services other than the
controllers until the Destination controller is modified to provide
identities via discovery.
The introduction of identity in 0626fa37 created new state in the
control plane's configuration that must be considered when re-installing
the control plane or when injecting pods.
This change alters `install` to fail if it would seem to conflict with
an existing installation. This behavior may be disabled with the
`--ignore-cluster` flag.
Furthermore, `inject` now _requires_ that it can fetch a configuration
from the control plane in order to operate. Otherwise the
`--ignore-cluster` and `--disable-identity` flags must be specified.
This change does not actually instrument pods to use identity yet---it
lays the framework for proxy identity without changing the test fixture
output (besides a change to how identity HA is configured).
Fixes#2531
https://github.com/linkerd/linkerd2/pull/2521 introduces an "Identity"
controller, but there is no way to include it in linkerd installation.
This change alters the `install` flow as follows:
- An Identity service is _always_ installed;
- Issuer credentials may be specified via the CLI;
- If no Issuer credentials are provided, they are generated each time `install` is called.
- Proxies are NOT configured to use the identity service.
- It's possible to override the credential generation logic---especially
for tests---via install options that can be configured via the CLI.
The new proxy has changed its configuration as follows:
- `LISTENER` urls are now `LISTEN_ADDR` addresses;
- `CONTROL_URL` is now `DESTINATION_SVC_ADDR`;
- `*_NAMESPACE` vars are no longer needed;
- The `PROXY_ID` is now the `DESTINATION_CONTEXT`;
- The "metrics" port is now the "admin" port, since it serves more than
just metrics;
- A readiness probe now checks a dedicated /ready endpoint eagerly.
Identity injection is **NOT** configured by this branch.
The proxy's TLS implementation has changed to use a new _Identity_ controller.
In preparation for this, the `--tls=optional` CLI flag has been removed
from install and inject; and the `ca` controller has been deleted. Metrics
and UI treatments for TLS have **not** been removed, as they will continue to
be valuable for the new Identity system.
With the removal of the old identity scheme, the Destination service's proxy
ID field is now set with an opaque string (e.g. `ns:emojivoto`) to enable
locality awareness.
* Defined the config annotations as new constants in labels.go
* Introduced the getOverride() functions to override configs
* Introduced new accessors to abstract with type casting
Signed-off-by: Ivan Sim <ivan@buoyant.io>
The CLI now specifies a default port, 50750, for the Linkerd dashboard.
If that port is not available, it resumes the original behavior of binding to a
free ephemeral port.
linkerd/linkerd2#1721 introduced a `--single-namespace` install flag,
enabling the control-plane to function within a single namespace. With
the introduction of ServiceProfiles, and upcoming identity changes, this
single namespace mode of operation is becoming less viable.
This change removes the `--single-namespace` install flag, and all
underlying support. The control-plane must have cluster-wide access to
operate.
A few related changes:
- Remove `--single-namespace` from `linkerd check`, this motivates
combining some check categories, as we can always assume cluster-wide
requirements.
- Simplify the `k8s.ResourceAuthz` API, as callers no longer need to
make a decision based on cluster-wide vs. namespace-wide access.
Components either have access, or they error out.
- Modify the web dashboard to always assume ServiceProfiles are enabled.
Reverts #1721
Part of #2337
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
It's sometimes helpful to spotcheck proxy metrics from a specific pod,
but doing so with kubectl requires a few steps.
Introduce a new `linkerd metrics` command. When given a pod name and
namespace, returns a dump of the proxy's /metrics endpoint.
Also modify the k8s.portforward module to accept initialized k8s config
and client objects, to enable testing.
Fixes#2350.
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
linkerd/linkerd2#2349 removed the `--single-namespace` flag, in favor of
runtime detection of cluster vs. namespace access, and also
ServiceProfile availability. This maintained control-plane support for
running in these two states.
This change requires control-plane components have cluster-wide
Kubernetes API access and ServiceProfile availability, and will error
out if not. Once #2349 merges, stage 1 install will be a requirement for
a successful stage 2 install.
Part of #2337
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Fixes#2377
In inject's ResourceConfig, renamed objMeta to podMeta since
it really points to the pod template metadata. And created a new field
workloadMeta that really points to the main workload (e.g. Deployment) metadata.
Refactored uninject to clean up the labels at both podMeta and
workloadMeta. Also it will remove all the labels and annotations that
start with "linkerd.io" except for the "linkerd.io/inject" annotation.
Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
The `linkerd-init` container requires the NET_ADMIN capability to modify
iptables. The `linkerd check` command was not verifying this.
Introduce a `has NET_ADMIN capability` check, which does the following:
1) Lists all available PodSecurityPolicies, if none found, returns
success
2) For each PodSecurityPolicy, validate one exists that:
- the user has `use` access AND
- provides `*` or `NET_ADMIN` capability
A couple limitations to this approach:
- It is testing whether the user running `linkerd check` has NET_ADMIN,
but during installation time it will be the `linkerd-init` pod that
requires NET_ADMIN.
- It assumes the presense of PodSecurityPolicies in the cluster means
the PodSecurityPolicy admission controller is installed. If the
admission controller is not installed, but PSPs exists that restrict
NET_ADMIN, `linkerd check` will incorrectly report the user does not
have that capability.
This PR also fixes the `can create CustomResourceDefinitions` check to
not specify a namespace when doing a `create` check, as CRDs are
cluster-wide.
Fixes#1732
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
- Created the pkg/inject package to hold the new injection shared lib.
- Extracted from `/cli/cmd/inject.go` and `/cli/cmd/inject_util.go`
the core methods doing the workload parsing and injection, and moved them into
`/pkg/inject/inject.go`. The CLI files should now deal only with
strictly CLI concerns, and applying the json patch returned by the new
lib.
- Proceeded analogously with `/cli/cmd/uninject.go` and
`/pkg/inject/uninject.go`.
- The `InjectReport` struct and helping methods were moved into
`/pkg/inject/report.go`
- Refactored webhook to use the new injection lib
- Removed linkerd-proxy-injector-sidecar-config ConfigMap
- Added the ability to add pod labels and annotations without having to
specify the already existing ones
Fixes#1748, #2289
Signed-off-by: Alejandro Pedraza <alejandro.pedraza@gmail.com>
As described in #2217, the controller returns TLS identities for results even
when the destination pod may not be able to participate in identity
requester: specifically, the other pod may not have the same controller
namespace or it may not be injected with identity.
This change introduces a new annotation, linkerd.io/identity-mode that is set
when injecting pods (via both CLI and webhook). This annotation is always
added.
The destination service now only returns TLS identities when this annotation
is set to optional on a pod and the destination pod uses the same controller.
These semantics are expected to change before the 2.3 release.
Fixes#2217
The control-plane components relied on a `--single-namespace` param,
passed from `linkerd install` into each individual component, to
determine which namespaces they were authorized to access, and whether
to support ServiceProfiles. This command-line flag was redundant given
the authorization rules encoded in the parent `linkerd install` output,
via [Cluster]Role[Binding]s.
Modify the control-plane components to query Kubernetes at startup to
determine which namespaces they are authorized to access, and whether
ServiceProfile support is available. This allows removal of the
`--single-namespace` flag on the components.
Also update `bin/test-cleanup` to cleanup the ServiceProfile CRD.
TODO:
- Remove `--single-namespace` flag on `linkerd install`, part of #2164
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Consolidate timeouts for `linkerd check`
- Moved the creation of contexts from inside the methods targeted by the
checks into a single place in the runCheck() and runCheckRPC() methods
where the context is built using a hard-coded timeout of 30 seconds.
- k8s' client-go doesn't allow passing along contexts, but it let's us
setting the Timeout manually.
- Reworded the description for the --wait option.
Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
Since 37ae423, deployments have been prefixed with linkerd-; however
the inject logic was not changed to take this into consideration when
constructing the controller's identity.
This means that the proxy's client to the control plane has been unable to
establish TLS'd communcation to the proxy-api. Previously, the proxy would
silently fall back to plaintext, but in master this behavior recently changed to
be stricter, so this bug will prevent the proxy from connecting to proxy-api
in any way.
* Add pod spec annotation to disable injection in CLI and auto-injector
* Remove support for linkerd.io/auto-inject label entirely
* Update based on review feedback
* Fix issue with finding the namespace of deployments applied to the default ns
Signed-off-by: Kevin Lingerfelt <kl@buoyant.io>
linkerd/website#105 introduced a FAQ page, providing resolutions for all
`linkerd check` failures.
Update each check to reference its corresponding section in the FAQ.
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Use `ca.NewCA()` for generating certs and keys for the proxy injector
- Remove from CA controller everything that dealt with the
webhook/proxy-injector
- Remove no longer needed proxy-injector volumes for 'trust-anchors' and
'webhook-secrets'
- Remove from the proxy-injector the retrieval of the trust anchor and
secrets
- tls flag during install is no longer needed for auto-inject to work
Fixes#2095 and fixes#2166
Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
DaemonSet stats are not currently shown in the cli stat command, web ui
or grafana dashboard. This commit adds daemonset support for stat.
Update stat command's help message to reference daemonsets.
Update the public-api to support stats for daemonsets.
Add tests for stat summary and api.
Add daemonset get/list/watch permissions to the linkerd-controller
cluster role that's created using the install command.
Update golden expectation test files for install command
yaml manifest output.
Update web UI with daemonsets
Update navigation, overview and pages to list daemonsets and the pods
associated to them.
Add daemonset paths to server, and ui apps.
Add grafana dashboard for daemonsets; a clone of the deployment
dashboard.
Update dependencies and dockerfile hashes
Add DaemonSet support to tap and top commands
Fixes of #2006
Signed-off-by: Zak Knill <zrjknill@gmail.com>
Add `linkerd uninject` command
uninject.go iterates through the resources annotations, labels,
initContainers and Containers, removing what we know was injected by
linkerd.
The biggest part of this commit is the refactoring of inject.go, to make
it more generic and reusable by uninject.
The idea is that in a following PR this functionality will get reused by
`linkerd inject` to uninject as as preliminary step to injection, as a
solution to #1970.
This was tested successfully on emojivoto with:
```
1) inject:
kubectl get -n emojivoto deployment -o yaml | bin/linkerd inject - |
kubectl apply -f -
2) uninject:
kubectl get -n emojivoto deployment -o yaml | bin/linkerd uninject - |
kubectl apply -f -
```
Also created unit tests for uninject.go. The fixture files from the inject
tests could be reused. But as now the input files act as outputs, they
represent existing resources and required these changes (that didn't
affect inject):
- Rearranged fields in alphabetical order.
- Added fields that are only relevant for existing resources (e.g.
creationTimestamp and status.replicas in StatefulSets)
Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
* Setup port-forwarding for linkerd dashboard command
* Output port-forward logs when --verbose flag is set
Signed-off-by: Kevin Lingerfelt <kl@buoyant.io>
Commit 1: Enable lint check for comments
Part of #217. Follow up from #1982 and #2018.
A subsequent commit will fix the ci failure.
Commit 2: Address all comment-related linter errors.
This change addresses all comment-related linter errors by doing the
following:
- Add comments to exported symbols
- Make some exported symbols private
- Recommend via TODOs that some exported symbols should should move or
be removed
This PR does not:
- Modify, move, or remove any code
- Modify existing comments
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Filtering by Kubernetes job was not supported. Also filtering by any unknown
type caused a panic.
Add filtering support by Kubernetes job, with special case mapping `job` to
`k8s_job`, to not conflict with Prometheus' job label.
Fix panic when unknown type specified as a `--from` or `--to` flag.
Fix `job` label from `linkerd-proxy` overwriting Prometheus `job` label at
collection time. This caused all metrics collected by proxy sidecars in
Kubernetes jobs to be collected into an incorrect Prometheus job, rather than
the expected `linkerd-proxy` Prometheus job.
Fix `unsupported resource type` tap error message incorrectly printing the
target resource rather than the destination.
Set `--controller-log-level debug` in `install_test.go` for easier debugging.
Expose `slow-cooker`'s metrics via a k8s service in the tap integration test, to
validate proxy requests with a job as destination.
Fixes#1872
Part of #627
Signed-off-by: Andrew Seigner <siggy@buoyant.io>
Andrew Seigner
Jack Price
Dennis Adjei-Baah
Alejandro Pedraza
Gaurav Kumar
Andrew Seigner
Katerina
Ivan Sim
Oliver Gould
Carol A. Scott
Tarun Pothulapati
Yan
Kevin Lingerfelt
zak
Radu M