Using sysctls in containers

Using sysctls in containers

Sysctl settings are exposed through Kubernetes, allowing users to modify certain kernel parameters at runtime. Only sysctls that are namespaced can be set independently on pods. If a sysctl is not namespaced, called node-level, you must use another method of setting the sysctl, such as by using the Node Tuning Operator.

Network sysctls are a special category of sysctl. Network sysctls include:

System-wide sysctls, for example net.ipv4.ip_local_port_range, that are valid for all networking. You can set these independently for each pod on a node.
Interface-specific sysctls, for example net.ipv4.conf.eth0.accept_local, that only apply to a specific interface. You cannot set these independently for each pod on a node. You set these by using a configuration in the tuning-cni after the network interfaces are created.

Moreover, only those sysctls considered safe are whitelisted by default; you can manually enable other unsafe sysctls on the node to be available to the user.

Additional resources

Node Tuning Operator

About sysctls

In Linux, the sysctl interface allows an administrator to modify kernel parameters at runtime. Parameters are available from the */proc/sys/* virtual process file system. The parameters cover various subsystems, such as:

kernel (common prefix: *kernel.*)
networking (common prefix: *net.*)
virtual memory (common prefix: *vm.*)
MDADM (common prefix: *dev.*)

More subsystems are described in Kernel documentation. To get a list of all parameters, run:

$ sudo sysctl -a

Namespaced and node-level sysctls

A number of sysctls are namespaced in the Linux kernels. This means that you can set them independently for each pod on a node. Being namespaced is a requirement for sysctls to be accessible in a pod context within Kubernetes.

The following sysctls are known to be namespaced:

*kernel.shm**
*kernel.msg**
*kernel.sem*
*fs.mqueue.**

Additionally, most of the sysctls in the net.* group are known to be namespaced. Their namespace adoption differs based on the kernel version and distributor.

Sysctls that are not namespaced are called node-level and must be set manually by the cluster administrator, either by means of the underlying Linux distribution of the nodes, such as by modifying the */etc/sysctls.conf* file, or by using a daemon set with privileged containers. You can use the Node Tuning Operator to set node-level sysctls.

Consider marking nodes with special sysctls as tainted. Only schedule pods onto them that need those sysctl settings. Use the taints and toleration feature to mark the nodes.

Safe and unsafe sysctls

Sysctls are grouped into safe and unsafe sysctls.

For system-wide sysctls to be considered safe, they must be namespaced. A namespaced sysctl ensures there is isolation between namespaces and therefore pods. If you set a sysctl for one pod it must not add any of the following:

Influence any other pod on the node
Harm the node health
Gain CPU or memory resources outside of the resource limits of a pod

Being namespaced alone is not sufficient for the sysctl to be considered safe.

Any sysctl that is not added to the allowed list on OKD is considered unsafe for OKD.

Unsafe sysctls are not allowed by default. For system-wide sysctls the cluster administrator must manually enable them on a per-node basis. Pods with disabled unsafe sysctls are scheduled but do not launch.

You cannot manually enable interface-specific unsafe sysctls.

OKD adds the following system-wide and interface-specific safe sysctls to an allowed safe list:

Table 1. System-wide safe sysctls
sysctl	Description
`kernel.shm_rmid_forced`	When set to `1`, all shared memory objects in current IPC namespace are automatically forced to use IPC_RMID. For more information, see shm_rmid_forced.
`net.ipv4.ip_local_port_range`	Defines the local port range that is used by TCP and UDP to choose the local port. The first number is the first port number, and the second number is the last local port number. If possible, it is better if these numbers have different parity (one even and one odd value). They must be greater than or equal to `ip_unprivileged_port_start`. The default values are `32768` and `60999` respectively. For more information, see ip_local_port_range.
`net.ipv4.tcp_syncookies`	When `net.ipv4.tcp_syncookies` is set, the kernel handles TCP SYN packets normally until the half-open connection queue is full, at which time, the SYN cookie functionality kicks in. This functionality allows the system to keep accepting valid connections, even if under a denial-of-service attack. For more information, see tcp_syncookies.
`net.ipv4.ping_group_range`	This restricts `ICMP_PROTO` datagram sockets to users in the group range. The default is `1 0`, meaning that nobody, not even root, can create ping sockets. For more information, see ping_group_range.
`net.ipv4.ip_unprivileged_port_start`	This defines the first unprivileged port in the network namespace. To disable all privileged ports, set this to `0`. Privileged ports must not overlap with the `ip_local_port_range`. For more information, see ip_unprivileged_port_start.

Table 2. Interface-specific safe sysctls
sysctl	Description
`net.ipv4.conf.IFNAME.accept_ra`	Accept IPv4 Router Advertisements; autoconfigure using them. It also determines whether or not to transmit router solicitations. Router solicitations are transmitted only if the functional setting is to accept router advertisements.
`net.ipv4.conf.IFNAME.accept_redirects`	Accept IPv4 ICMP redirect messages.
`net.ipv4.conf.IFNAME.accept_source_route`	Accept IPv4 packets with strict source route (SRR) option.
`net.ipv4.conf.IFNAME.arp_accept`	Define behavior for gratuitous ARP frames with an IPv4 address that is not already present in the ARP table: `0` - Do not create new entries in the ARP table. `1` - Create new entries in the ARP table.
`net.ipv4.conf.IFNAME.arp_notify`	Define mode for notification of IPv4 address and device changes.
`net.ipv4.conf.IFNAME.disable_policy`	Disable IPSEC policy (SPD) for this IPv4 interface.
`net.ipv4.conf.IFNAME.secure_redirects`	Accept ICMP redirect messages only to gateways listed in the interface’s current gateway list.
`net.ipv4.conf.IFNAME.send_redirects`	Send redirects is enabled only if the node acts as a router. That is, a host should not send an ICMP redirect message. It is used by routers to notify the host about a better routing path that is available for a particular destination.
`net.ipv6.conf.IFNAME.accept_ra`	Accept IPv6 Router advertisements; autoconfigure using them. It also determines whether or not to transmit router solicitations. Router solicitations are transmitted only if the functional setting is to accept router advertisements.
`net.ipv6.conf.IFNAME.accept_redirects`	Accept IPv6 ICMP redirect messages.
`net.ipv6.conf.IFNAME.accept_source_route`	Accept IPv6 packets with SRR option.
`net.ipv6.conf.IFNAME.arp_accept`	Define behavior for gratuitous ARP frames with an IPv6 address that is not already present in the ARP table: `0` - Do not create new entries in the ARP table. `1` - Create new entries in the ARP table.
`net.ipv6.conf.IFNAME.arp_notify`	Define mode for notification of IPv6 address and device changes.
`net.ipv6.neigh.IFNAME.base_reachable_time_ms`	This parameter controls the hardware address to IP mapping lifetime in the neighbour table for IPv6.
`net.ipv6.neigh.IFNAME.retrans_time_ms`	Set the retransmit timer for neighbor discovery messages.

The interface name is represented by the IFNAME token, and is replaced with the actual name of the interface at runtime.

Additional resources

Linux networking documentation

Starting a pod with safe sysctls

You can set sysctls on pods using the pod’s securityContext. The securityContext applies to all containers in the same pod.

Safe sysctls are allowed by default.

This example uses the pod securityContext to set the following safe sysctls:

kernel.shm_rmid_forced
net.ipv4.ip_local_port_range
net.ipv4.tcp_syncookies
net.ipv4.ping_group_range

To avoid destabilizing your operating system, modify sysctl parameters only after you understand their effects.

Use this procedure to start a pod with the configured sysctl settings.

In most cases you modify an existing pod definition and add the securityContext spec.

Procedure

Create a YAML file sysctl_pod.yaml that defines an example pod and add the securityContext spec, as shown in the following example:

apiVersion: v1
kind: Pod
metadata:
  name: sysctl-example
  namespace: default
spec:
  containers:
  - name: podexample
    image: centos
    command: ["bin/bash", "-c", "sleep INF"]
    securityContext:
      runAsUser: 2000 (1)
      runAsGroup: 3000 (2)
      allowPrivilegeEscalation: false (3)
      capabilities: (4)
        drop: ["ALL"]
  securityContext:
    runAsNonRoot: true (5)
    seccompProfile: (6)
      type: RuntimeDefault
    sysctls:
    - name: kernel.shm_rmid_forced
      value: "1"
    - name: net.ipv4.ip_local_port_range
      value: "32770       60666"
    - name: net.ipv4.tcp_syncookies
      value: "0"
    - name: net.ipv4.ping_group_range
      value: "0           200000000"

1	`runAsUser` controls which user ID the container is run with.
2	`runAsGroup` controls which primary group ID the containers is run with.
3	`allowPrivilegeEscalation` determines if a pod can request to allow privilege escalation. If unspecified, it defaults to true. This boolean directly controls whether the `no_new_privs` flag gets set on the container process.
4	`capabilities` permit privileged actions without giving full root access. This policy ensures all capabilities are dropped from the pod.
5	`runAsNonRoot: true` requires that the container will run with a user with any UID other than 0.
6	`RuntimeDefault` enables the default seccomp profile for a pod or container workload.

Create the pod by running the following command:
```
$ oc apply -f sysctl_pod.yaml
```

Verify that the pod is created by running the following command:

$ oc get pod

Example output

NAME              READY   STATUS            RESTARTS   AGE
sysctl-example    1/1     Running           0          14s

Log in to the pod by running the following command:
```
$ oc rsh sysctl-example
```
Verify the values of the configured sysctl flags. For example, find the value kernel.shm_rmid_forced by running the following command:
```
sh-4.4# sysctl kernel.shm_rmid_forced
```
Expected output
```
kernel.shm_rmid_forced = 1
```

Starting a pod with unsafe sysctls

A pod with unsafe sysctls fails to launch on any node unless the cluster administrator explicitly enables unsafe sysctls for that node. As with node-level sysctls, use the taints and toleration feature or labels on nodes to schedule those pods onto the right nodes.

The following example uses the pod securityContext to set a safe sysctl kernel.shm_rmid_forced and two unsafe sysctls, net.core.somaxconn and kernel.msgmax. There is no distinction between safe and unsafe sysctls in the specification.

To avoid destabilizing your operating system, modify sysctl parameters only after you understand their effects.

The following example illustrates what happens when you add safe and unsafe sysctls to a pod specification:

Procedure

Create a YAML file sysctl-example-unsafe.yaml that defines an example pod and add the securityContext specification, as shown in the following example:

apiVersion: v1
kind: Pod
metadata:
  name: sysctl-example-unsafe
spec:
  containers:
  - name: podexample
    image: centos
    command: ["bin/bash", "-c", "sleep INF"]
    securityContext:
      runAsUser: 2000
      runAsGroup: 3000
      allowPrivilegeEscalation: false
      capabilities:
        drop: ["ALL"]
  securityContext:
    runAsNonRoot: true
    seccompProfile:
      type: RuntimeDefault
    sysctls:
    - name: kernel.shm_rmid_forced
      value: "0"
    - name: net.core.somaxconn
      value: "1024"
    - name: kernel.msgmax
      value: "65536"

Create the pod using the following command:
```
$ oc apply -f sysctl-example-unsafe.yaml
```

Verify that the pod is scheduled but does not deploy because unsafe sysctls are not allowed for the node using the following command:

$ oc get pod

Example output

NAME                       READY             STATUS            RESTARTS   AGE
sysctl-example-unsafe      0/1               SysctlForbidden   0          14s

Enabling unsafe sysctls

A cluster administrator can allow certain unsafe sysctls for very special situations such as high performance or real-time application tuning.

If you want to use unsafe sysctls, a cluster administrator must enable them individually for a specific type of node. The sysctls must be namespaced.

You can further control which sysctls can be set in pods by specifying lists of sysctls or sysctl patterns in the forbiddenSysctls and allowedUnsafeSysctls fields of the Security Context Constraints.

The forbiddenSysctls option excludes specific sysctls.
The allowedUnsafeSysctls option controls specific needs such as high performance or real-time application tuning.

Due to their nature of being unsafe, the use of unsafe sysctls is at-your-own-risk and can lead to severe problems, such as improper behavior of containers, resource shortage, or breaking a node.

Procedure

List existing MachineConfig objects for your OKD cluster to decide how to label your machine config by running the following command:

$ oc get machineconfigpool

Example output

NAME     CONFIG                                             UPDATED   UPDATING   DEGRADED   MACHINECOUNT   READYMACHINECOUNT   UPDATEDMACHINECOUNT   DEGRADEDMACHINECOUNT   AGE
master   rendered-master-bfb92f0cd1684e54d8e234ab7423cc96   True      False      False      3              3                   3                     0                      42m
worker   rendered-worker-21b6cb9a0f8919c88caf39db80ac1fce   True      False      False      3              3                   3                     0                      42m

Add a label to the machine config pool where the containers with the unsafe sysctls will run by running the following command:
```
$ oc label machineconfigpool worker custom-kubelet=sysctl
```

Create a YAML file set-sysctl-worker.yaml that defines a KubeletConfig custom resource (CR):

apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: custom-kubelet
spec:
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: sysctl (1)
  kubeletConfig:
    allowedUnsafeSysctls: (2)
      - "kernel.msg*"
      - "net.core.somaxconn"

1	Specify the label from the machine config pool.
2	List the unsafe sysctls you want to allow.

Create the object by running the following command:
```
$ oc apply -f set-sysctl-worker.yaml
```

Wait for the Machine Config Operator to generate the new rendered configuration and apply it to the machines by running the following command:

$ oc get machineconfigpool worker -w

After some minutes the UPDATING status changes from True to False:

NAME     CONFIG                                             UPDATED   UPDATING   DEGRADED   MACHINECOUNT   READYMACHINECOUNT   UPDATEDMACHINECOUNT   DEGRADEDMACHINECOUNT   AGE
worker   rendered-worker-f1704a00fc6f30d3a7de9a15fd68a800   False     True       False      3              2                   2                     0                      71m
worker   rendered-worker-f1704a00fc6f30d3a7de9a15fd68a800   False     True       False      3              2                   3                     0                      72m
worker   rendered-worker-0188658afe1f3a183ec8c4f14186f4d5   True      False      False      3              3                   3                     0                      72m

Create a YAML file sysctl-example-safe-unsafe.yaml that defines an example pod and add the securityContext spec, as shown in the following example:

apiVersion: v1
kind: Pod
metadata:
  name: sysctl-example-safe-unsafe
spec:
  containers:
  - name: podexample
    image: centos
    command: ["bin/bash", "-c", "sleep INF"]
    securityContext:
      runAsUser: 2000
      runAsGroup: 3000
      allowPrivilegeEscalation: false
      capabilities:
        drop: ["ALL"]
  securityContext:
    runAsNonRoot: true
    seccompProfile:
      type: RuntimeDefault
    sysctls:
    - name: kernel.shm_rmid_forced
      value: "0"
    - name: net.core.somaxconn
      value: "1024"
    - name: kernel.msgmax
      value: "65536"

Create the pod by running the following command:

$ oc apply -f sysctl-example-safe-unsafe.yaml

Expected output

Warning: would violate PodSecurity "restricted:latest": forbidden sysctls (net.core.somaxconn, kernel.msgmax)
pod/sysctl-example-safe-unsafe created

Verify that the pod is created by running the following command:

$ oc get pod

Example output

NAME                         READY   STATUS    RESTARTS   AGE
sysctl-example-safe-unsafe   1/1     Running   0          19s

Log in to the pod by running the following command:
```
$ oc rsh sysctl-example-safe-unsafe
```
Verify the values of the configured sysctl flags. For example, find the value net.core.somaxconn by running the following command:
```
sh-4.4# sysctl net.core.somaxconn
```
Expected output
```
net.core.somaxconn = 1024
```

The unsafe sysctl is now allowed and the value is set as defined in the securityContext spec of the updated pod specification.

Additional resources

Setting interface-level network sysctls