Repository navigation
Releases
v0.79.0
Behavior Discovery names running processes by the path they were started by
A process that was already running when Tetragon started used to be listed by the file the kernel loaded for it, with links followed: a sleep that is a link to coreutils showed as coreutils, /bin/sh as busybox or dash. A process rule matches the path a process is started by, so a whitelist built from those names refused the real ones the next time the pod started. They are now listed as /usr/bin/sleep and /bin/sh.
Processes that were already running are also learned when Tetragon is installed or restarted while Sentinel is running, or a node joins, within a minute.
New permission: exec in Tetragon's namespace
Tetragon's API does not carry the started-by path, so Sentinel reads it on the node by running one fixed script in the tetragon container. deploy/sentinel.yaml adds a Role and RoleBinding, sentinel-tetragon-exec, in kube-system for that. Kubernetes cannot grant exec for one kind of pod only, so this is exec in every pod of that namespace: someone who takes over Sentinel can get onto the nodes with it. To do without:
kubectl -n kube-system delete rolebinding,role sentinel-tetragon-execEverything else keeps working, and those processes are listed as before. Details: docs/install.md.
Upgrading
Apply the manifest again, not only the new image, or the Role is missing and the names stay as they were:
kubectl apply -f https://raw.githubusercontent.com/cooloo9871/K8s_Sentinel/v0.79.0/deploy/sentinel.yamlNames learned before the upgrade stay in a profile until it is cleared. Where Tetragon is installed outside kube-system, create the Role and binding in its namespace.
Images
# K8s Sentinel v0.79.0: every image an offline install needs.
# Tetragon chart 1.8.0, Trivy Operator chart 0.38.1.
ghcr.io/cooloo9871/sentinel:v0.79.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
# artifacts
# Trivy's vulnerability databases. They are OCI artifacts, not container
# images: docker pull cannot fetch them; skopeo and oras can. They change
# daily, and a scan is as current as the copy in your registry.
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
v0.78.0
CRI-O nodes work with Tetragon's default settings
On CRI-O, Tetragon names no pod for the processes of a container, so Behavior Discovery learned nothing and Security Events had no namespace or pod. Sentinel now works out the pod itself, from the container runtime that started each process. Processes that were already running are covered too. Nothing changes on containerd.
Upgrading from v0.77.0 on CRI-O: switch two Tetragon settings back
v0.77.0's installers switched on tetragon.cri.enabled and tetragon.cgidmap.enabled for CRI-O nodes. With those on, a Tracing Policy that names a namespace or a pod selector blocks nothing. This release no longer sets them. If your cluster has them on, switch them off:
helm repo update cilium
helm -n kube-system upgrade tetragon cilium/tetragon --version <installed version> --reuse-values \
--set tetragon.cri.enabled=false \
--set tetragon.cgidmap.enabled=false
kubectl -n kube-system rollout status ds/tetragon
kubectl -n sentinel-system rollout restart deploy/sentinelhelm -n kube-system list shows the installed version. For an offline install, apply manifests/tetragon.yaml from this bundle. Details: docs/install.md.
Images
# K8s Sentinel v0.78.0: every image an offline install needs.
# Tetragon chart 1.8.0, Trivy Operator chart 0.38.1.
ghcr.io/cooloo9871/sentinel:v0.78.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
# artifacts
# Trivy's vulnerability databases. They are OCI artifacts, not container
# images: docker pull cannot fetch them; skopeo and oras can. They change
# daily, and a scan is as current as the copy in your registry.
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
v0.77.0
Added
- The installers set Tetragon up for CRI-O. On CRI-O, Tetragon cannot tell which pod a process belongs to from the name of its cgroup: Behavior Discovery lists only
runc, Security Events carry no pod and no Quarantine this pod button.deploy/install.sh,deploy/install-job.yamland the offline installer now look at the container runtime the nodes report and, when every node runs CRI-O, have Tetragon ask the runtime through its socket (/var/run/crio/crio.sock). The offline bundle carries a second manifest,tetragon-crio.yaml, for this; the images are the same.
A Tetragon that is already installed is left as it is, and so is a cluster where only some nodes run CRI-O. Nodes on CRI-O has the commands to set it by hand.
Not yet run end to end on a CRI-O cluster.
Offline install
sentinel-offline-v0.77.0.tar.gz holds the manifests and scripts; sentinel-images-v0.77.0.txt lists every image to mirror:
# K8s Sentinel v0.77.0: every image an offline install needs.
# Tetragon chart 1.8.0, Trivy Operator chart 0.38.1.
ghcr.io/cooloo9871/sentinel:v0.77.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
# artifacts
# Trivy's vulnerability databases. They are OCI artifacts, not container
# images: docker pull cannot fetch them; skopeo and oras can. They change
# daily, and a scan is as current as the copy in your registry.
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
v0.76.1
Fixed
- Hubble Relay in a cluster with a custom cluster domain. Sentinel looked for the relay at
hubble-relay.<ns>.svc.cluster.local:80, which does not resolve when the cluster domain is anything else, and the Network Topology stayed empty withname resolver error: produced zero addresses. The default is nowhubble-relay.<ns>.svc:80, completed by the pod's DNS search path whatever the domain is.HUBBLE_RELAY_ADDRESSstill overrides it.
Offline install
sentinel-offline-v0.76.1.tar.gz holds the manifests and scripts; sentinel-images-v0.76.1.txt lists every image to mirror:
# K8s Sentinel v0.76.1: every image an offline install needs.
# Tetragon chart 1.8.0, Trivy Operator chart 0.38.1.
ghcr.io/cooloo9871/sentinel:v0.76.1
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
# artifacts
# Trivy's vulnerability databases. They are OCI artifacts, not container
# images: docker pull cannot fetch them; skopeo and oras can. They change
# daily, and a scan is as current as the copy in your registry.
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
v0.76.0
Quarantine from the Network Topology
Click a pod on the Network Topology and its detail panel now offers Quarantine this pod, the same confirmed action a Security Event has. Use it for a pod seen connecting out, or being reached from outside, before any event names it.
- Asks first, naming the namespace and pod
- The node turns red with a lock at once, and the panel says it is quarantined
- Admins only; external addresses and nodes cannot be quarantined
- Release from Policies → Quarantine, as before
Images
# K8s Sentinel v0.76.0: every image an offline install needs.
# Tetragon chart 1.8.0, Trivy Operator chart 0.38.1.
ghcr.io/cooloo9871/sentinel:v0.76.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
# artifacts
# Trivy's vulnerability databases. They are OCI artifacts, not container
# images: docker pull cannot fetch them; skopeo and oras can. They change
# daily, and a scan is as current as the copy in your registry.
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
Offline install: download sentinel-offline-v0.76.0.tar.gz and see Offline install.
v0.75.0
Offline install
- Fixed: on a cluster without Trivy Operator,
install.shended withno matches for kind "ClusterComplianceReport"and stopped before waiting for the operator. The CRDs are now applied first and waited for. An install that ended this way is complete in everything Sentinel uses; nothing needs to be run again.
Images
ghcr.io/cooloo9871/sentinel:v0.75.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
The last two are Trivy's vulnerability databases: OCI artifacts.
Upgrading
No manifest change from v0.70.0.
v0.74.0
Offline install
mirror-images.shneeds onlydockerorpodman, and says which it uses. It no longer looks for skopeo, oras or nerdctl on the machine.- Trivy's vulnerability databases, which neither docker nor podman can carry as images, go through skopeo run from a container image (
quay.io/skopeo/stable:v1.18).savetakes that image along, sopushhas it offline. It is a tool of the script and is not pushed to your registry. - Log in to the registry before
push(docker login,podman login).
Images
ghcr.io/cooloo9871/sentinel:v0.74.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
The last two are Trivy's vulnerability databases: OCI artifacts.
Upgrading
No manifest change from v0.70.0.
v0.73.0
Offline install
mirror-images.sh push <registry> --insecure-registry: push to a registry on plain HTTP, or with a certificate the pushing machine does not trust (x509: certificate signed by unknown authority).INSECURE=1in the environment still does the same.- The help of
install.shno longer ends in a line of code.
Images
ghcr.io/cooloo9871/sentinel:v0.73.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
The last two are Trivy's vulnerability databases: OCI artifacts, copied with skopeo or oras.
Upgrading
No manifest change from v0.70.0.
v0.72.0
Offline install
install.sh --insecure-registry: for a private registry whose certificate is signed by a CA of your own, or that serves plain HTTP. Trivy then downloads its databases, and reads the images it scans, without verifying certificates. The nodes' container runtime still has to trust the registry by its own configuration.TRIVY_INSECURE=trueon Sentinel does the same for its registry scans.
Images
ghcr.io/cooloo9871/sentinel:v0.72.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
The last two are Trivy's vulnerability databases: OCI artifacts, copied with skopeo or oras.
Upgrading
No manifest change from v0.70.0.
v0.71.0
Offline install
For clusters with no internet access. See Offline install.
sentinel-offline-v0.71.0.tar.gz: manifests and scripts; needs only kubectl on the cluster sidesentinel-images-v0.71.0.txt: every image this release needs
Every release from this one on carries both.
Images
ghcr.io/cooloo9871/sentinel:v0.71.0
mirror.gcr.io/aquasec/trivy:0.76.1
mirror.gcr.io/aquasec/trivy-operator:0.36.1
quay.io/cilium/hubble-export-stdout:v1.1.1
quay.io/cilium/tetragon-operator:v1.8.0
quay.io/cilium/tetragon:v1.8.0
mirror.gcr.io/aquasec/trivy-db:2
mirror.gcr.io/aquasec/trivy-java-db:1
The last two are Trivy's vulnerability databases: OCI artifacts, copied with skopeo or oras.
Upgrading
No manifest change from v0.70.0.