Case: Defender is not being deployed due to insufficient resources
Memory limit to 512M — If the Defender exceeds it’s memory limit, it will be terminated. The kubelet will restart it as with any other type of runtime feature.
Requesting 256m CPU — This leads to blocking. Meaning if the node does not have this amount of available allocation then the the scheduler will refuse to place the Pod. Though in this case the Defender will not be killed for excessive CPU usage. Consider a scenario where the customer is using a node with 1 CPU. If he runs an application that requests for 600 milli CPUs and the Defender that requests for ~300 milli CPUs, both containers utilizing CPU to the most, then the extra 100 milli CPUs will be distributed to the app and the defender in a 2:1 ratio.
To be reviewed Most customers can’t simply give themselves more permissions (like we can in our lab environments). The proper way to resolve this is:
Disable service account monitoring.
List the min perms required, so that the user can ask for them from the team that manages the cluster.
Prisma Cloud lets you monitor service accounts and the permissions granted to them for each resource in your Kubernetes or OpenShift clusters. When you install a Defender DaemonSet, you can enable service account monitoring. To do so, you must be a cluster admin.
If you get the following error when installing a Defender DaemonSet, then you don’t have the right permissions.
Insufficient permissions to monitor service account. Please disable service account monitoring or elevate your account:
You have two options:
Disable service account monitoring in the Defender DaemonSet install command.
Grant yourself cluster-admin privileges.
To grant yourself cluster-admin privileges:
$ kubectl create clusterrolebinding cluster-admin-binding \ --clusterrole cluster-admin \ --user $(gcloud config get-value account)