* types: fix usage of strings.Split() for parsing CNI_ARGS
strings.Split() returns a slice, in this case with two elements of
the key and value. As such we shouldn't range over the slice when
the code is expecting a 2-element slice of key/value.
Otherwise we get errors for valid CNI_ARGS like:
2022-04-22T11:53:54Z [error] CreateCNIRuntimeConf: CNI_ARGS K8S_POD_NAMESPACE=openshift-etcd K8S_POD_NAMESPACE 17 is not recognized as CNI arg, skipped
2022-04-22T11:53:54Z [error] CreateCNIRuntimeConf: CNI_ARGS K8S_POD_NAMESPACE=openshift-etcd openshift-etcd 14 is not recognized as CNI arg, skipped
Fixes: d7d2a99ab5 ("Replace setenv with runtimeConfig set")
Signed-off-by: Dan Williams <dcbw@redhat.com>
* types/conf: fix handling of CNI_ARGS additions to rt.Args
We want to set the CNI_ARGS value in rt.Args if the existing value
is empty, not if the key doesn't exist yet. Since the rt.Args array
is pre-created with the K8S args keys, empty values of those keys
couldn't be overwritten with the previous scheme that just checked
if the key existed.
If the CNI_ARGS key wasn't found in rt.Args then add it; previously
a typo ("isExists != false") prevented that with inverted logic.
Signed-off-by: Dan Williams <dcbw@redhat.com>
* multus: entrypoint: disallow incompatible cni versions
When top level CNI version is 0.4.0 or more, nested CNI version
can't be less than 0.4.0 since these are incompatible. This
closes issue #737.
Signed-off-by: Balazs Nemeth <bnemeth@redhat.com>
* multus: thick: disallow incompatible cni versions
Similarly to disallowing incompatible versions in entrypoint.sh,
add the same logic in go for the thick plugin.
Signed-off-by: Balazs Nemeth <bnemeth@redhat.com>
* multus: add unit test for incompatible cni versions
Signed-off-by: Balazs Nemeth <bnemeth@redhat.com>
setenv refers environment variables, which is unique in process,
not unique to go routine. Hence it may causes some issue in multi
threaded case, hence it is replaced with libcni's runtimeConfig
value set to set these variables at libcni side, after process
fork.
To keep consistency between actual network and CNI result in cache,
update libcni cache when multus add/del default routes by
`default-route` network selection.
The test was just checking that a READ/WRITE fsnotify.Event for
the multus configuration was being seen; this patch changes this
behavior, and assures that the delegateCNI configuration update results
in turn on the update of the multus configuration file.
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
Using a unique path for the config generation will lessen the
amount of events caught by the test that checks if the multus
configuration must be re-generated as a result a default cluster
configuration update.
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* build: install the multus binary in an init container
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* build: generate kubeconfig via go
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* build: generate multus cni configuration via golang
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* build: provide a docker img for daemon based deployments
We will have 2 different images (only on amd64 archs):
- legacy entrypoint script based
- daemonized process
The `image-build` docker action is updated, to build these 2 images.
There will be 2 different deployment specs, along with e2e test
lanes, one for each of the aforementioned alternatives.
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* build: delegate CNI config watch loop via golang
For the thick-plugin alternative, provide the watch loop for
configuration regeneration via a golang binary.
Over time, this binary is expected to run the control loop to watch
out for pod updates.
To enable current multus users to chose when they upgrade to this new
deployment setup, these changes are provided in separate multus images,
having a different yaml spec files. Both of these alternatives are
tested e2e, since a new lane is introduced.
The following libraries are introduced, along with the motivation for
adding them:
- dproxy: allows traversing the default network configuration arbitrarily,
similar to what an X path / JSON path tool provides.
Repo is available at [0].
- fsnotify: watch for changes in the default CNI configuration file.
Repo is available at [1].
The config map providing the default network CNI configuration is not
copied over, since originally, the user was not required to install a
default network CNI plugin first, but, nowadays, this is a required
step of multus.
As such, it is no longer required to provide a default CNI
configuration.
[0] - https://github.com/koron/go-dproxy
[1] - https://github.com/fsnotify/fsnotify
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* run gofmt
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* refactor: make the builder pattern more idiomatic to golang
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
* build: update github actions to release new imgs
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
Because the tests fake out the OS exec functionality, the environment that
would usually be passed to a plugin instead gets passed as a string array
to the ExecPlugin() function in the tests. But when verifying the expected
environment this function was looking at the actual OS environment rather
than the passed string array. Fix that.
Use that to test various cases of the pod UID being passed to plugins.
Signed-off-by: Dan Williams <dcbw@redhat.com>
If the runtime passes a pod UID via K8S_POD_UID (which both CRIO and
containerd do as of mid-2021) then fail if the pod we get from the
Kube API has a different UID. This would indicate that the pod was
deleted and recreated while Multus was attempting to set up
networking for the old pod instance's sandbox, and it's pointless
to continue setting up a sandbox for a dead pod instance.
Also pass the pod UID through to plugins so they can perform
additional checking and validation on the pods they get from the
Kube API.
Signed-off-by: Dan Williams <dcbw@redhat.com>
Multus community decide that multus support the Kubernetes that
Kubernetes community supports, so changed vendor code to the
oldest version of currently supported.
Thus not repeating ourselves, nor shadowing the global variables
that had the same name (pollInterval / pollTimeout).
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
Also retry on the following error types (cmd ADD/DEL):
- IsInternalError
- IsConnectionReset
- IsConnectionRefused
Signed-off-by: Miguel Duarte Barroso <mdbarroso@redhat.com>
The case withot DeviceInfo/ResourceName, we could skip functions
related to device info. This change skips to call these functions
if delegates does not have DeviceInfo/ResourceName.
Today, Multus will always log via STDERR, and these logs will then
logged by the Kubelet. If we also choose to have Multus log to a file by
setting the LogFile option in the CNI configuration, the same logs will
be logged twice.
This commit provide the option to disable logging to STDERR.
Signed-off-by: Yun Zhou <yunz@nvidia.com>
When we call mergeRuntimeConfig, the global RuntimeConfig gets
overwritten with the result of the merging, thus affecting the
subsequent delegates.
Do not modify the global RuntimeConfig and instead make a copy
when merging it.
Also, if a value has been provided for CNIDeviceInfoFile in the
delegate's runtimeconfig, overwrite it to avoid possible name
colissions.
Signed-off-by: Adrian Moreno <amorenoz@redhat.com>
