kubevirt / hyperconverged-cluster-operator / 16857722445

76%

push

github

Deploy wasp-agent using HCO (#3682)

* improve the RBAC cluster role generic operands

currently there is no need to calculate each time the
desired state of the RBAC resource because its independent
of HCO CR changes, therefore now it stores a constant desired
state of the resource.

Signed-off-by: Igor Bezukh <ibezukh@redhat.com>

* introduce the wasp-agent operand

wasp-agent is a node-local agent that is responsible to mimic
the kubernetes swap feature in out-of-tree manner. It will be used
until we will develop an operational approach to use kuberentes swap

wasp-agent is composed of the following components
- Daemonset that deploys the node-local agent on the workker nodes
- Service account
- Cluster role and cluster role binding
- Security Context Constraints

wasp-agent deployment is conditional, the deployment is being
triggered as far as the memory overcommit percentage is higher
than 100%. In that case in order to make the memory overcommit safer
wasp-agent allows memory tiering with swap so that workload idle
memory would be able to swap instead of OOMKilled.

wasp-agent needs access to pods in all namespaces, therefore
it is required to use the cluster role.

since wasp-agent mimics the kubelet functionality, it is required
to use the privileged SCC in order to be able to function properly on
the node (configuring the cgroups, talk to CRI-O, etc.)

wasp-agent needs to watch all the running pods in the cluster,
therefore it uses the allow-access-cluster-services network
policy in order to communicate with the API server.

wasp-agent currently supported only on OpenShift.

Signed-off-by: Igor Bezukh <ibezukh@redhat.com>

---------

Signed-off-by: Igor Bezukh <ibezukh@redhat.com>

279 of 313 new or added lines in 7 files covered. (89.14%)

7511 of 9803 relevant lines covered (76.62%)

1.82 hits per line