This commit enhances the Kilo agent internal HTTP server to include
pprof endpoints. For simplicity, this commit migrates the internal
server creation to https://github.com/metalmatze/signal/internalserver,
which allows for easy registration of common internal server
observability endpoints.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Currently, Kilo _only_ supports adding firewall rules via the legacy
iptables API. This means that on systems using nftables in the host
network namespace, the namespace will be polluted and both firewall
infrastructures will be used, causing unexpected and difficult
to predict interactions. In other words, networking may not work as
expected on nftables-based systems.
This PR fixes this by using the iptables-wrappers project [0] to install
run-time detection of the in-use iptables backend.
[0] https://github.com/kubernetes-sigs/iptables-wrappers
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit updates the CI configuration for Kilo to ensure that the
documentation, specifically the generated docs, are up-to-date.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
* CNI: bump to 1.0.1
This commit bumps the declared version of CNI in the Kilo manifests to
1.0.1. This is possible with no changes to the configuration lists
because our simple configuration is not affected by any of the
deprecations, and there was effectively no change between 0.4.0 and
1.0.0, other than the declaration of a stable API. Similarly, this
commit also bumps the version of the CNI library and the plugins
package.
Bumping to CNI 1.0.0 will help ensure that Kilo stays compatible with
container runtimes in the future.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
* vendor: revendor
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Currently,the job to build kgctl binaries is named `linux`, which
suggests to the reader that the job is only building binaries for Linux,
when it is in fact building binaries for Linux, Darwin, and Windows.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Currently, CI only runs for PRs to the main branch. This commit modifies
the configuration so that it runs for PRs to any branch.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Currently, when rendering the configuration for a Peer, the allowed
location configs of any segment are erroneously ignored, meaning that an
administrator will have to manually edit the configuration to get the
expected behavior from a Peer. This commit fixes the generation of the
configuration.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Right now, the persistent keepalive field of the Peer CRD is always
interpretted as nanoseconds and not seconds. This causes a mismatch
between Kilo's expected behavior and the actual interval that is given
to Peers. Because the interval is interpretted as nanoseconds the value
rounds down to 0 seconds.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Currently, when a node is behind NAT, it is possible that routes to the
node's private IP address, i.e. routes necessary to communicate with the
Kubelet and any Pods on the host network, will not be created because
the private IP is seen as the same as the location's endpoint and is
thus skipped because trying to encapsulate traffic to the endpoint would
break communiation with the endpoint itself.
This logic is not correct for nodes that are behind NAT, because the
endpoin that the node reports may not be the same as the discovered
endpoint for the location. Instead, we should compare the private IP
address to the discovered endpoint.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
After running make, the licenses for the generated go files is updated,
resulting in a diff in the repository. This makes later invocations of
`make container` generate tags with `$ARCH-$SHA-dirty` rather than just
`$ARM-$SHA`, which causes `make manifest` to fail, as some of the
images cannot be found.
A more permanent fix would be to ensure that running `make container`
does not cause the go code to unnecessarily regenerated, but this will
at least fix CI until next year.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit enhances the iptables controller to disable reconciliation
of IPv6 rules whenever it detects that IPv6 is disabled in the kernel,
in order to fix#259.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit adds a stage to the GitHub Actions CI workflow to verify
that vendor and go.mod/sum are always up-to-date. If the vendored files
require any changes then CI will fail. This ensures that the repo
remains the source of truth.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
When updating Kilo to the latest version of the CustomResourceDefinition
API, the Kilo Peer CRD was incorrectly scoped as a namespaced resource
due to differences in the ergonomics of the tooling.
This commit fixes the scoping of the Peer CRD to be cluster-wide.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Since Kilo now uses the `kilo.squat.ai/discovered-endpoints` annotation
for Peer discovery, Kilo no longer needs to update Peer resources, so we
can remove this permission from the ClusterRole. Note, the RBAC in the
manifests is not used today, but we eventually want to migrate to this.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit standardizes how we display a `note` in our docs.
Previously, we used a mix of `_Note_:`, `__Note:__`, `> Note`, and `>
**Note**:` among others. Now, all notes appear as `> **Note**:`.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit adds a new test suite to the e2e tests that validates
multi-cluster connectivity in Kilo. This is really just an extension of
the testing of the Peers CRD and related tooling that also exercises the
`--allowed-ips` flag of the `kgctl` tool.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit enables the filtering of e2e tests that should be run when
using the Makefile's `e2e` target through the specification of the
`BASH_UNIT_FLAGS` environment variable. The value of this variable will
be passed as arguments to the `bash_unit` command, enabling filtering of
tests.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit simplifies the check_adjacency helper to us the curl_pod
helper rather than a re-written version of it. It also simplifies the
curl_pod helper slightly to avoid the need for an additional shell.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit allows the kind cluster configuration to be parameterized at
call time. This enables the test suite to build multiple clusters with
different configurations, e.g. different CIDRs, different numbers of
nodes, etc.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit modifies the logic of the check_adjacent helper so that
rather than expecting the argument to be n^2+n it expects simply n. This
makes it easier to update the caller when the number of nodes in the
cluster changes.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit modifies the e2e shell scripts so that the KUBECONFIG
variable does not need to be exported. This will become important once
we allow the e2e test suite to launch multiple clusters, e.g. to test
multi-cluster connectivity.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Currently, each test suite spins up its own kind cluster, which results
in longer e2e test times as each test suite needs to wait for the
cluster to be ready and for images to download. This commit creates two
new virtual test suites that are run before and after the actual e2e
tests and are responsible for creating and destroying a kind cluster
respectively. Any test suite that needs a fresh cluster can still spin
up its own using the `create_cluster` helper in the lib.sh file.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
We seem to be running into
https://github.com/kubernetes-sigs/kind/issues/2240: kube-proxy is
crashlooping, which in turn causes CoreDNS to fail to connect to the API
server over the Service CIDR, which in turn cases DNS resolution to fail
in the cluster, which in turn causes the e2e tests to fail.
To address this,
this commit bumps the kind version to 0.11.1.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit splits the e2e tests into different suites that all reuse
helper functions from the lib.sh file. This split ensures that the tests
within a suite can all have the same setup, e.g. all tests depending on
a full-mesh will be set up with a full-mesh, and is resilient against
changes in test ordering.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit adds a new e2e test fot the recently introduced
allowed-location-ips annotation. This test annotates the control-plane
node with an allowed IP and then ensures this IPs is reachable from
the curl helper Pod, which is now guaranteed to be scheduled on a
worker node.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit adds an e2e test that ensures that invalid Kilo Peers are
rejected thanks to the improved OpenAPI v3 validation.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
The recent test refactor wrapped the adjacency check in a `retry`,
however, the `check_adjacent` function had an `assert` in it, which
meant that the test would fail immediately and not be retried. This
commit moves the assertion out of the check, which also allows for
improved code reuse.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit adds better handling of the default kubeconfig location in
the kgctl binary for cases where the `$KUBECONFIG` environment variable
is not set. In these cases, kgctl will default to
`$HOME/.kube/config`, putting it in line with tools like `kubectl` and
`kind`.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit adds testing of the VPN feature of Kilo to the e2e tests.
Also, in order to make the tests less flaky, this commit ensures that
the Kilo Pods use the "root" kubeconfig to connect to the API rather
than the Kubernetes API's cluster IP, which can become unavailable when
networking is reconfigured.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit refactors the e2e tests in the following ways:
* bump the version of `bash_unit` to 1.7.2: the previous version was
from 2018;
* remove the unused `block` function;
* fix the order of `block_unitil_ready_by_name` and `check_ping`: we
should not attempt to interact with pods before they are considered
ready;
* extract the retry logic into a reusable function called `retry`; and
* retry the `check_adjacent` function for robustness.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit bumps the Golang toolchain version used to build Kilo to go
1.16 to support builing kgctl for the arm64 M1 macs.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit fixes lint warnings produced by shellcheck. This simplifies
some of the `if`s and protects against variables with spaces. We may
want to consider shellchecking all scripts in the repo.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit cleans up some typos in the documentation, clarifies some
explanations, fixes a bash script that would not expand a variable, and
renames the API documentation generator command from gen-docs to
docs-gen to match the output binary name.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
This commit updates the instructions for installing Kilo. It also fixes
the title-casing of a section on the README and makes a visual change to
the userspace doc to try to make the note clearer.
Signed-off-by: Lucas Servén Marín <lserven@gmail.com>
Lucas Servén Marín