ocpnetsplit package

ocpnetsplit.machineconfig module

This module generates MachineConfig to deploy network-split systemd units, which implements the network split functionality.

References:

ocpnetsplit.machineconfig.create_file_dict(basename, content, target_dir='/etc')

Create Ignition config spec for given file basename and content, to be used in a MachineConfig spec. File will be placed given target_dir, but note that MCO can only change files in /etc and /var directories.

Parameters:

  • basename (str) – basename of the file

  • content (str) – content of the file

  • target_dir (str) – absolute path where to place the file, eg. /etc

Returns:

Ignition storage file config spec

Return type:

dict

Raises:

ValueError – if given basename or target_dir is invalid

ocpnetsplit.machineconfig.create_latency_mc_dict(role, latency, latency_spec=None)

Create MachineConfig dict with latency systemd units and scripts.

Parameters:

  • mcp (string) – name of MachineConfig role (and also MachineConfigPool) where the MachineConfig generated by this function should be deployed. Usually master or worker.

  • latency (int) – default zone latency created via Linux Traffic Control in ms

  • ( (latency_spec) – py:class`ocpnetsplit.zone.ZoneLatSpec`): specific latency between given zones (optional).

Returns:

MachineConfig dict

Return type:

dict

ocpnetsplit.machineconfig.create_script_dict(script_name)

Create file dict with given shell script from ocpnetsplit module.

Parameters:

script_name (string) – name of the shell script

Returns:

Ignition storage file config spec

Return type:

dict

ocpnetsplit.machineconfig.create_split_mc_dict(role)

Create MachineConfig dict with network-split systemd units and scripts.

Parameters:

mcp (string) – name of MachineConfig role (and also MachineConfigPool) where the MachineConfig generated by this function should be deployed. Usually master or worker.

Returns:

MachineConfig dict

Return type:

dict

ocpnetsplit.machineconfig.create_systemdunit_dict(unit_filename)

Create file dict with given systemd unit file from ocpnetsplit module.

Parameters:

unit_filename (string) – name of the systemd unit file

Returns:

Ignition storage file config spec

Return type:

dict

ocpnetsplit.machineconfig.create_unit_dict(name, content)

Create Ignition config spec for given systemd unit name and content, to be used in a MachineConfig spec.

Parameters:

  • name (str) – name of systemd unit

  • content (str) – content of the file

Returns:

Ignition systemd unit config spec

Return type:

dict

ocpnetsplit.machineconfig.create_zone_mc_dict(role, zone_env)

Create MachineConfig dict with network zone config env file.

Parameters:

  • mcp (string) – name of MachineConfig role (and also MachineConfigPool) where the MachineConfig generated by this function should be deployed. Usually master or worker.

  • zone_env (string) – content of network-split.env file with zone configuration, as created by ocpnetsplit.zone.ZoneConfig.get_env_file()

Returns:

MachineConfig dict

Return type:

dict

ocpnetsplit.machineconfig.get_new_mc(role, name_suffix, priority=99)

Initialize new (almost empty) MachineConfig dict.

Parameters:

  • role (string) – name of MachineConfig role

  • name_suffix (string) – suffix of resulting MachineConfig name

ocpnetsplit.main module

Module with a public API of ocp-network-split project. One can either use the command line tools (as implemented via main functions in this module), or to use the python functions defined here directly.

ocpnetsplit.main.check_split(nodes, split_name, use_ssh=False)

Checks status of split via systemctl list-timers on all nodes of the cluster.

Parameters:

  • nodes (list) – list of all nodes from all zones

  • split_name (str) – network split configuration specification, eg. ab, see ocpnetsplit.zone.NETWORK_SPLITS constant

  • use_ssh (bool) – if true, connect to the nodes via ssh; use oc debug node otherwise

Raises:

ValueError – when invalid split_name is specified

ocpnetsplit.main.get_networksplit_mc_spec(zone_env=None, split=False, latency=0, latency_spec=None)

Create MachineConfig spec to install network split firewall tweaking script and unit files on all cluster nodes.

Parameters:

  • zone_env (str) – content of firewall zone env file specifying node ip addresses for each cluster zone, as created by ocpnetsplit.zone.ZoneConfig.get_env_file()

  • split (bool) – when true, support for net splits will be included

  • latency (int) – default zone latency created via Linux Traffic Control in ms, when the value is zero, support for latency is not included

  • ( (latency_spec) – py:class`ocpnetsplit.zone.ZoneLatSpec`): specific latency between given zones (optional).

Returns:

list of dictionaries with MachineConfig spec

Return type:

machineconfig_spec

ocpnetsplit.main.get_zone_config(zone_a, zone_b, zone_c, zone_x_addrs=None)

For each valid ocp-network-split zone name (see ocpnetsplit.zone.ZONES), translate it’s given topology.kubernetes.io/zone label into list of ip addresses of all nodes in the zone.

Parameters:

  • zone_a (str) – value of zone a label

  • zone_b (str) – value of zone b label

  • zone_c (str) – value of zone c label

  • zone_x_addrs (list) – list of ip addresses in external zone x

Returns:

object with list of node ip addresses for each zone name

ocp network split works with (a, b, …), see ocpnetsplit.zone.ZONES).

Return type:

ZoneConfig

ocpnetsplit.main.get_zone_config_fromfile(file_content, translate_hostname=True)

Get zone config from ini file, which contains node fqdn entries for each zone.

ocpnetsplit.main.main_multisetup()

Simple multi cluster version of command line interface to generate MachineConfig yaml and env file to deploy on OCP/Ceph clusters.

Example usage:

$ ocp-network-split-multisetup zones.ini --mc mc.yaml --env network-split.env
ocpnetsplit.main.main_sched()

Simple command line interface to schedule given cluster network split.

Example usage:

$ ocp-network-split-sched ab-bc -t 2021-03-18T18:45 --split-len 30
$ ocp-network-split-sched ab-bc
ocpnetsplit.main.main_setup()

Simple command line interface to generate MachineConfig yaml to deploy to make scheduling network splits possible.

Example usage:

$ ocp-network-split-setup -a arbiter -b d1 -c d2 -o mc.yaml
$ oc create -f mc.yaml
$ oc get mcp
ocpnetsplit.main.run_ssh_node(cmd_list, node, timeout=600)

Run given command on given node via ssh assuming connection details like username and keys are specified via ~/.ssh/config file.

Parameters:

  • cmd_list (list) – a command to run, eg. ["uname", "-a"] will execute uname -a process on the node

  • node (str) – hostname of k8s node where to execute the command

  • timeout (int) – command timeout specified in seconds, optional

Returns:

ssh stdout, ssh souterr

Return type:

tuple

ocpnetsplit.main.schedule_split(nodes, split_name, target_dt, target_length, use_ssh=False)

Schedule start and stop of network split on all nodes of the cluster.

Parameters:

  • nodes (list) – list of all nodes from all zones

  • split_name (str) – network split configuration specification, eg. ab, see ocpnetsplit.zone.NETWORK_SPLITS constant

  • target_dt (datetime) – requested start time of the network split

  • target_length (int) – number of minutes specifying how long the network split configuration should be active

  • use_ssh (bool) – if true, connect to the nodes via ssh; use oc debug node otherwise

Raises:

ValueError – in case invalid split_name or target_dt is specified.

ocpnetsplit.ocp module

ocpnetsplit.ocp.get_all_node_ip_addrs(node, kubeconfig=None, oc_executable=None)

Get all ip addresses (both internal and external) of given node.

Parameters:

  • node (str) – name of OCP node

  • kubeconfig (str) – file path to kubeconfig (optional, use only if you need to override the default)

  • oc_executable (str) – file path of oc command (optional, use only if you need to override the default)

Returns:

node ip addressess (as strings)

Return type:

list

ocpnetsplit.ocp.list_cluster_nodes(zone_name=None, kubeconfig=None, oc_executable=None)

Get cluster nodes of a whole cluster or from given zone only.

Parameters:

  • zone_name (str) – name of k8s topology zone to list nodes within, if not specified, nodes from whole cluster will be listed

  • kubeconfig (str) – file path to kubeconfig (optional, use only if you need to override the default)

  • oc_executable (str) – file path of oc command (optional, use only if you need to override the default)

Returns:

node ip addressess (as strings)

Return type:

list

ocpnetsplit.ocp.run_oc(cmd_list, kubeconfig=None, oc_executable=None, timeout=600)

Run given oc command and log all it’s output.

Parameters:

  • cmd_list (list) – oc command to run, eg. ["get", "nodes"] will execute oc get nodes process

  • timeout (int) – command timeout specified in seconds, optional

  • kubeconfig (str) – file path to kubeconfig (optional, use only if you need to override the default)

  • oc_executable (str) – file path of oc command (optional, use only if you need to override the default)

Returns:

stdout, stderr of the command executed

Return type:

tuple

ocpnetsplit.ocp.run_oc_debug_node(cmd_list, node, kubeconfig=None, oc_executable=None)

Run given command on given node via oc debug node.

Parameters:

  • cmd_list (list) – a command to run, eg. ["uname", "-a"] will execute uname -a process on the node

  • node (str) – name of k8s node where to execute the command, with or without node/ prefix

  • kubeconfig (str) – file path to kubeconfig (optional, use only if you need to override the default)

  • oc_executable (str) – file path of oc command (optional, use only if you need to override the default)

Returns:

cmd_out (combined stdout and stderr of the executed command),

oc_out (output from oc debug process itself)

Return type:

tuple

ocpnetsplit.zone module

ocpnetsplit.zone.NETWORK_SPLITS = ('ab', 'bc', 'ab-bc', 'ab-ac', 'ax', 'ax-bx-cx')

Available network split configurations. For every valid network split value, there is a systemd timer unit named network-split-{split}-setup@.timer. Network split configuration consists of list of zone tuples, where each zone tuple represents a disrupted zone connection.

ocpnetsplit.zone.ZONES = ('a', 'b', 'c', 'x')

Stable zone identifiers as defined and used by ocp-network-split.

class ocpnetsplit.zone.ZoneConfig

Bases: object

ZoneConfig is tracking ip addresses of nodes in each cluster zone.

add_node(zone, node)

Add a node ip address into a zone.

Parameters:

  • zone (str) – zone identification (one of ZONES)

  • node (str) – ip address of a node

add_nodes(zone, nodes)

Add list of node ip addresses into a zone.

Parameters:

  • zone (str) – zone identification (one of ZONES)

  • nodes (list) – list of string representation of node ip addresses

get_env_file()

Generate content of env file for firewall script.

Returns:

content of firewall environment file with zone configuration

Return type:

str

get_nodes(zone=None)

Return set of node ip addresses in given zone.

Parameters:

  • zone (str) – zone identification (one of ZONES), if not

  • specified

  • be (zone filtering is not applied and all nodes will)

  • returned

Returns:

string representation of node ip addresses of given zone

Return type:

list

class ocpnetsplit.zone.ZoneLatSpec(**kwargs)

Bases: object

Describe latency values between given zones.

Validation of input latency spec is necessary to catch mistakes as early as possible (debugging the problem later on a live cluster increases cost of debugging and a fix significantly).

get_cli_args()

Generate command line arguments for network-latency.sh script representing latency spec of this object.

load_arguments(latency_spec)

Load latency spec from the given list produced by argparse parser.

Parameters:

latency_spec (list) – List of latency specs from argument parser. Eg.: ['ab=10', 'bc=10'].

load_dict(latency_spec)

Load latency spec from the given dict.

Parameters:

  • latency_spec (dict) – specific latency between given zones, for

  • zones (example {'ab'=11} will represent 22ms RTT latency between)

  • b. (a and)