Datafiles
This section provides the users with the schema for each datafile. For additional information about how to use the datafiles described in this page, please refer to the getting started guide.
Note
The format for the regex keys is as follows:
(<some regex>)
For example: (.*upti.*)
would match any sentence which has upti
inside it.
Markup in Datafiles
For every datafile below you can use the markup syntax as seen here to allow variable substitution and references similar to the Django template language.
Note
Your testbed datafile values are available using this syntax:
%{testbed.<path>.<path>}
Verification Datafile
# Verification Datafile schema
# ----------------------------
#
# Production schema with commentary from the devs
extends: # Use this field to extend an existing yaml Verification file,
# allowing you to create an inheritance hierarchy.
# Supports full path/names or name of file in the same dir.
# ``Genie`` provides a Master set of verifications for you to
# extend from as a start. It is found at:
# <$virtual_env>/lib/python3.4/site-packages/genie/infra/verifications.yaml
variables: # Use this field to store any information as dict to be shared
# within this datafile
# Can be retrieved : %{variables.<field>}
parameters: # same purpose with variables
# Can be retrieved : %{parameters.<field>}
uids: # Mention which verifications to execute
groups: # Mention which verification group to execute
<verification name>: # Verification name goes here
source: # This section is used to let
# Genie know where the verification class
# is located.
pkg: # Abstraction package.
# (Optional if no abstraction)
# (Example: genie_libs)
class: # Location of the class that contains the verification.
# (Example: ops.ospf.ospf.Ospf)
processors: # This section is used to let
# Genie know where the pyATS processors are located
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre-processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
groups: # Execution group for this verification.
# (Optional)
count: # How many times to execute this verification.
# (Default: 1)
# (Optional)
exclude: # Attributes to be ignored.
# (Optional)
# (Supports regex)
iteration: # Tells Genie to rerun the verification in case
# the result is not valid.
# (Optional)
attempt: # Number of iteration attempts.
# (Optional)
# (Default: 1)
interval: # How long to sleep between rerun, in seconds.
# (Optional)
# (Default: 0)
devices: # Devices list to execute the verification.
# Can either be an alias or device hostname
# as defined in the pyats testbed file.
# If there is no device, the verification will not execute.
devices_attributes: # attributes for the devices
# (Optional)
<device name>: # Name of the device; must be same as devices list
iteration: # Tells Genie to rerun the verification in case
# the result is not valid.
# (Optional)
attempt: # Number of iteration attempts.
# (Optional)
# (Default: 1)
interval: # How long to sleep between rerun, in seconds.
# (Optional)
# (Default: 0)
# Any extra key will be passed as a parameters to the verification
<key>: <value>
parameters: # (Optional if no parser parameters)
# Any key will be passed as a parser parameters
<key>: <value>
# Any extra key will be passed as a parameters to the verification
<key>: <value>
global_processors: # Section to let Genie where the pyATS processors is located
# These processors will run for all verifications
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre-processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
Trigger Datafile
# Trigger Datafile schema
# -----------------------
#
# Production schema with commentary from the devs
extends: # Use this field to extend an existing yaml Trigger file,
# allowing you to create an inheritance hierarchy.
# Supports full path/names or names of file in the the same dir.
# ``Genie`` provides a master set of triggers for you to extend
# from as a start. It is found at:
# <$virtual_env>/lib/python3.4/site-packages/genie/infra/triggers.yaml
variables: # Use this field to store any information as dict to be shared
# within this datafile
# Can be retrieved : %{variables.<field>}
parameters: # same purpose with variables
# Can be retrieved : %{parameters.<field>}
uids: # Mention which triggers to execute
groups: # Mention which trigger group to execute
<trigger name>: &trigger_arguments # Trigger name goes here
source: # This section is used to let
# Genie know where the trigger class
# is located.
pkg: # Abstraction package.
# (optional if no abstraction)
# (Example: genie_libs)
class: # Location of the class that contains the trigger.
# (Example: sdk.triggers.shutnoshut.shutnoshut.TriggerShutNoShutOspf)
timeout: # (Optional) This section is used for timeout of trigger
max_time: # (Optional) Maximum wait time for the trigger. in second.
interval: # (Optional) Wait time between iteration when looping is needed. in second.
processors: # This section is used to let
# Genie know where the pyATS processors is located
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre-processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
order: # accepts a list of processors
# (Optional if no special ordering expected)
# This list will decide the execution order of processors
# any processors not defined here will be excluded
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
order: # accepts a list of processors
# (Optional if no special ordering expected)
# This list will decide the execution order of processors
# any processors not defined here will be excluded
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
order: # accepts a list of processors
# (Optional if no special ordering expected)
# This list will decide the execution order of processors
# any processors not defined here will be excluded
groups: # Execution group for this trigger.
# (Optional)
count: # How many times to execute this trigger.
# (Default: 1)
# (Optional)
devices: # Devices lists to execute the verification.
# Can either be an alias or device hostname
# as defined in the pyats testbed file.
# If there is no device, the verification will not execute.
devices_attributes: # attributes for the devices
# (Optional)
<device name>: # Name of the device; must be same as devices list
# Any extra key will be passed as a parameter to the verification
<key>: <value>
verifications: # Local verification to execute
<verification name>: # Verification name
# It must match one of the names of a
# existing verification in
# the verification_datafile.
devices: # Devices lists to execute the verification.
# Can either be an alias or device hostname
# as defined in the pyats testbed file.
# If there is no device, the verification will not execute.
devices_attributes: # attributes for the devices
# (Optional)
<device name>: # Name of the device; must be same as devices list
iteration: # Tells Genie to rerun the verification in case
# the result is not valid.
# (Optional)
attempt: # How many times to rerun.
# (Optional)
# (Default: 1)
interval: # How long to sleep between rerun, in seconds.
# (Optional)
# (Default: 0)
iteration: # Tells Genie to rerun the verification in case
# the result is not valid.
# (Optional)
attempt: # How many times to rerun.
# (Optional)
# (Default: 1)
interval: # How long to sleep between rerun, in seconds.
# (Optional)
# (Default: 0)
parameters: # (Optional if no parser parameters)
# Any key will be passed as a parser parameters
<key>: <value>
sections: # Adding processor to some Trigger section
<section name>: # Name of the section to add processor to
parameters: # Optional
<key>: <value> # Adding in the parameters key along with
# key/values will allow you to pass in arguments
# on a per-section basis. Arguments defined in
# the main body of the trigger will be overridden
# by these arguments
processors: # This section is used to let
# Genie know where the pyATS processors is located
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre-processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
static: # Most trigger learn dynamically certain data.
# This key allows to specify which value to use instead of learning.
# (Optional will learn all the keys dynamically in the trigger)
<key>: <value>
num_values: # Most trigger learn dynamically certain data.
# This key allows to specify how many value of each type should be learnt
# (Optional by default will used the one in the trigger code)
# (Example:
# interface 1
# vlan: 'all'
<key>: <value>
skip_global_verifications: # Skip global verifications
# Accepts a list
# Any extra key will be passed as a parameter to the trigger.
<key>: <value>
global_processors: # Section to let Genie where the pyATS processors is located
# These processors will run for all triggers
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre-processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
order: # accepts a list of triggers
- <trigger name>
# (Optional if no special ordering expected)
# This list will decide the execution order of triggers
# any trigger not defined here will be excluded
# OR
# A list of dictionaries for each trigger
# with trigger arguments supported as above
- <trigger name>: *trigger_arguments
Configuration Datafile
# Config Datafile schema
# -----------------------
#
# Production schema with commentary from the devs
extends: # Use this field to extend an existing yaml Trigger file,
# allowing you to create an inheritance hierarchy.
# Supports full path/names or names of file in the the same dir.
# (Optional)
variables: # Use this field to store any information as dict to be shared
# within this datafile
# Can be retrieved : %{variables.<field>}
parameters: # same purpose with variables
# Can be retrieved : %{parameters.<field>}
devices: # Devices list to apply the configuration
<device name>: # Name of the device.
# Can either be an alias or device hostname
# as defined in the pyats testbed file.
<Number>: # This number will decide the order of
# the applied configurations on the device.
type: # Type of configuration (setup or cleanup).
# (Optional, defaults to 'setup')
config: # Full path of config file to apply on the device.
# (Optional)
jinja2_config: # Full path of jinja2 config file
# (Optional)
jinja2_arguments: # Arguments to be added into the jinja2 file
# (Optional)
<key>: <value> # Key/Value pair of arguments
configure_arguments: # Arguments passed to unicon configure() service
<key>: <value> # Key/Value pair of arguments
sleep: # As device configurations take some time to
# stabilize, this sleep will tell Genie
# how long to wait before continuing, in seconds.
# (Optional)
# (Default: 0)
invalid: # When applying configuration, we might see
# some errors or warnings. Any error or warning
# patterns specified in the `invalid` key will
# fail the configure subsection.
# (Optional)
# (Supports regex)
exclude_config_check: # Configuration to be excluded
# in the check_config subsection.
# (Optional)
# (Supports regex)
# (Example: ['(.*uptime.*),]
PTS Datafile
# PTS Datafile schema
# -------------------
#
# Production schema with commentary from the devs
extends: # Use this field to extend an existing yaml PTS file,
# allowing you to create an inheritance hierarchy.
# Supports full path/names or names of file in the the same dir.
# ``Genie`` provides a master set of PTS for you to extend
# from as a start. It may be found at:
# <$virtual_env>/lib/python3.4/site-packages/genie/infra/pts.yaml
variables: # Use this field to store any information as dict to be shared
# within this datafile
# Can be retrieved : %{variables.<field>}
parameters: # same purpose with variables
# Can be retrieved : %{parameters.<field>}
<feature name>: # Feature name to profile goes here.
source: # This section is used to let
# Genie know where the profile class
# is located.
pkg: # Abstraction package.
# (optional if no abstraction)
# (Example: genie_libs)
class: # Location of the class.
# (Example: ops.ospf.ospf.Ospf)
exclude: # Attributes to be ignored.
# (Optional)
# (Supports regex)
devices: # Devices list to execute the verification.
# Can either be an alias or device hostname
# as defined in the pyats testbed file.
# If there is no device, the verification will not execute.
devices_attributes: # attributes for the devices
# (Optional)
<device name>: # Name of the device; must be same as devices list
# Any extra key will be passed as a parameter to the verification
<key>: <value>
Mapping Datafile
Mapping datafile is not mandatory in Genie, it is required when user desires to control the connection per device.
By Default (if not provided);
Genie
will connect all the devices in the testbed yaml file as soon as
they have either;
A single connection defined
connections:
defaults:
class: 'unicon.Unicon'
a:
protocol: telnet
ip: "100.100.100.100"
port: 1234
A connection named
cli
within provided multiple connections
connections:
defaults:
class: 'unicon.Unicon'
a:
protocol: telnet
ip: "100.100.100.100"
port: 1234
cli:
protocol: telnet
ip: "200.200.200.200"
port: 5678
# Mapping Datafile schema
# -----------------------
#
# Production schema with commentary from the devs
variables: # Use this field to store any information as dict to be shared
# within this datafile
# Can be retrieved : %{variables.<field>}
parameters: # same purpose with variables
# Can be retrieved : %{parameters.<field>}
devices: # Devices
<device name>: # Name of the device.
# Can either be an alias or device hostname
# as defined in the pyats testbed file.
context: # Context to priotize for the Genie execution.
pool_size: # Connection pool size, in case of using pyATS connection
# pool feature.
# Optional, `harness` will issue single connection if it
# is not provided.
label: # One device in the testbed must be designated as the `uut`
# for ``Genie``. If no device is already named `uut`, then
# you must map it via `label` field, as shown in the example
# above. In the event that the device is already named uut
# via `alias` in the testbed file, then there is no need to
# use label field.
mapping: #
cli: # Which connection to use for Cli commands.
yang: # Which connection to use for Yang commands.
xml: # Which connection to use for xml commands.
rest: # Which connection to use for rest commands.
topology: # Topologies
links: # Links
<link name>: # Name of the link
label: # Label to override the alias of the specified link
<device name>: # Name of the device for interfaces
interfaces: # The device's interfaces
<interface name>: # Name of the interface
label: # Label to override th alias of the specified interface
Different connections using Mapping datafile:
The user can connect with different connections using the following schema.
Note
: The user can either use this new schema or the one mentioned above.
Note
: via and alias keywords are mandatory when using new mapping schema.
devices:
uut:
mapping:
context:
cli:
- via: cli
alias: cli
pool_size: 5
sleep: 3
- via: ssh
alias: cli2
yang:
- via: yang
alias: netconf
# New Mapping Datafile schema for different connections
# -------------------------------------------------
variables: # Use this field to store any information as dict to be shared
# within this datafile
# Can be retrieved : %{variables.<field>}
parameters: # same purpose with variables
# Can be retrieved : %{parameters.<field>}
devices: # Devices
<device name>: # Name of the device.
# Can either be an alias or device hostname
# as defined in the pyats testbed file.
label: # One device in the testbed must be designated as the `uut`
# for ``Genie``. If no device is already named `uut`, then
# you must map it via `label` field, as shown in the example
# above. In the event that the device is already named uut
# via `alias` in the testbed file, then there is no need to
# use label field.
mapping: # To map which connection to use for in the context.
context: # Context for the Genie execution.
via: #Specify the connection to use. Eg. cli, ssh, yang etc
alias: #The alias name of the connection. Eg. my_connection
pool_size: # Connection pool size, in case of using pyATS connection
# pool feature.
# Optional, `harness` will issue single connection if it
# is not provided.
sleep: #Take a nap after making connection. By default 5 seconds for yang, gnmi, restconf,
netconf.
Subsection datafile
# Subsection Datafile schema
# --------------------------
#
# Production schema with commentary from the devs
extends: # Use this field to extend an existing yaml Subsection data file,
# allowing you to create an inheritance hierarchy.
# Supports full path/names or name of file in the the same dir.
variables: # Use this field to store any information as dict to be shared
# within this datafile
# Can be retrieved : %{variables.<field>}
parameters: # same purpose with variables
# Can be retrieved : %{parameters.<field>}
setup: # CommonSetup subsections
# (optional if common setup)
sections: # Subsections
# (Optional if no new subsection)
<subsection name>: # Name of the Subsection.
method: # Location of the subsection methods
# (Example: genie_libs.sdk.libs.subsections.my_subsection)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
processors: # This section is used to let
# Genie know where the pyATS processors is located
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre-processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
order: # accepts a list of subsections
# (Optional if no special ordering expected)
# This list will decide the execution order of subsections
# any subsection not defined here will be excluded
processors: # Processors
# (Optional; only if want processor for common_setup)
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
cleanup: # CommonSetup subsections
# (optional if common cleanup)
sections: # Subsections
# (Optional if no new subsection)
<subsection name>: # Name of the Subsection.
method: # Location of the subsection methods
# (Example: genie_libs.sdk.libs.subsections.my_subsection)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
processors: # This section is used to let
# Genie know where the pyATS processors is located
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre-processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
order: # accepts a list of subsections
# (Optional if no special ordering expected)
# This list will decide the execution order of subsections
# any subsection not defined here will be excluded
processors: # Processors
# (Optional; only if want processor for common_cleanup)
pre: # pre processors
# (Optional if no pre processors)
# Accepts a list of pre processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
post: # post processors
# (Optional if no post processors)
# Accepts a list of post processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>
exception: # exception processors
# (Optional if no exception processors)
# Accepts a list of exception processors
# (Example: [sdk.libs.prepostprocessor.sleep_processor])
# or use abstraction and define optional parameters
<processor name>: # Processor name goes here
pkg: # Abstraction package
# (Example: genie_libs)
method: # Location of the processor method
# (Example: sdk.libs.prepostprocessor.sleep_processor)
parameters: # (Optional if no parameters)
# Any key will be passed as a parameters
<key>: <value>