NETCONF testtool

NETCONF testtool is a set of standalone runnable jars that can:

  • Simulate NETCONF devices (suitable for scale testing)

  • Stress/Performance test NETCONF devices

  • Stress/Performance test RESTCONF devices

These jars are part of OpenDaylight’s controller project and are built from the NETCONF codebase in OpenDaylight.

Nexus contains 3 executable tools:

  • executable.jar - device simulator

  • stress.client.tar.gz - NETCONF stress/performance measuring tool

  • perf-client.jar - RESTCONF stress/performance measuring tool

Tip

Each executable tool provides help. Just invoke java -jar <name-of-the-tool.jar> --help

NETCONF device simulator

NETCONF testtool (or NETCONF device simulator) is a tool that

  • Simulates 1 or more NETCONF devices

  • Is suitable for scale, performance or crud testing

  • Uses core implementation of NETCONF server from OpenDaylight

  • Allows for automatic sending of connector configurations through RESTCONF to the controller. This makes it easy for the OpenDaylight distribution (Karaf) to connect to all simulated devices once the simulator starts.

  • Provides broad configuration options

  • Can start a fully fledged MD-SAL datastore

  • Supports notifications

Building testtool

  1. Check out latest NETCONF repository from git

  2. Move into the opendaylight/netconf/tools/netconf-testtool/ folder

  3. Build testtool using the mvn clean install command

Downloading testtool

Netconf-testtool is now part of default maven build profile for controller and can be also downloaded from nexus. The executable jars for testtool can be found by release at this parent directory: nexus-artifacts

Running testtool

  1. After successfully building or downloading, move into the opendaylight/netconf/tools/netconf-testtool/target/ folder and there is file netconf-testtool-[VERSION]-executable.jar (or if downloaded from nexus just take that *-executable.jar file)

  2. Execute this file using, e.g.:

    java -jar netconf-testtool-[VERSION]-executable.jar

    This execution runs the testtool with default for all parameters and you should see this log output from the testtool :

    10:31:08.206 [main] INFO  o.o.c.n.t.t.NetconfDeviceSimulator - Starting 1, SSH simulated devices starting on port 17830
10:31:08.675 [main] INFO  o.o.c.n.t.t.NetconfDeviceSimulator - All simulated devices started successfully from port 17830 to 17830

Default Parameters

The default parameters for testtool are:

  • Use SSH

  • Run 1 simulated device

  • Device port is 17830

  • YANG modules used by device are only: ietf-netconf-monitoring, ietf-yang-types, ietf-inet-types (these modules are required for device in order to support NETCONF monitoring and are included in the netconf-testtool)

  • Connection timeout is set to 30 minutes (quite high, but when testing with 10000 devices it might take some time for all of them to fully establish a connection)

  • Debug level is set to false

  • Other default parameters can be seen with --help option

Verifying testtool

To verify that the simulated device is up and running, we can try to connect to it using command line ssh tool. Execute this command to connect to the device:

ssh admin@localhost -p 17830 -s netconf

Just accept the server with yes (if required) and provide any password (testtool accepts all users with all passwords). You should see the hello message sent by simulated device.

Testtool help

usage: netconf testtool [-h] [--edit-content EDIT-CONTENT] [--async-requests {true,false}]
                        [--thread-amount THREAD-AMOUNT] [--throttle THROTTLE]
                        [--controller-auth-username CONTROLLER-AUTH-USERNAME]
                        [--controller-auth-password CONTROLLER-AUTH-PASSWORD] [--controller-ip CONTROLLER-IP]
                        [--controller-port CONTROLLER-PORT] [--device-count DEVICES-COUNT]
                        [--devices-per-port DEVICES-PER-PORT] [--schemas-dir SCHEMAS-DIR]
                        [--notification-file NOTIFICATION-FILE]
                        [--initial-config-xml-file INITIAL-CONFIG-XML-FILE] [--starting-port STARTING-PORT]
                        [--generate-config-connection-timeout GENERATE-CONFIG-CONNECTION-TIMEOUT]
                        [--generate-config-address GENERATE-CONFIG-ADDRESS]
                        [--generate-configs-batch-size GENERATE-CONFIGS-BATCH-SIZE]
                        [--distribution-folder DISTRO-FOLDER] [--ssh {true,false}] [--exi {true,false}]
                        [--debug {true,false}] [--md-sal {true,false}] [--time-out TIME-OUT] [--ip IP]
                        [--thread-pool-size THREAD-POOL-SIZE] [--rpc-config RPC-CONFIG]

netconf testtool

named arguments:
  -h, --help             show this help message and exit
  --edit-content EDIT-CONTENT
  --async-requests {true,false}
                         (default: false)
  --thread-amount THREAD-AMOUNT
                         The number of threads to use for configuring devices. (default: 1)
  --throttle THROTTLE    Maximum amount of async requests that  can  be  open  at  a time, with mutltiple threads
                         this gets divided among all threads (default: 5000)
  --controller-auth-username CONTROLLER-AUTH-USERNAME
                         Username for HTTP basic authentication to destination controller. (default: admin)
  --controller-auth-password CONTROLLER-AUTH-PASSWORD
                         Password for HTTP basic authentication to destination controller. (default: admin)
  --controller-ip CONTROLLER-IP
                         Ip of controller if  available  it  will  be  used  for  spawning netconf connectors via
                         topology   configuration   as   a    part   of   URI(http://<controller-ip>:<controller-
                         port>/rests/data/...) otherwise it  will  just  start  simulated  devices  and  skip the
                         execution of PATCH requests
  --controller-port CONTROLLER-PORT
                         Port of controller if available  it  will  be  used  for spawning netconf connectors via
                         topology   configuration   as   a    part   of   URI(http://<controller-ip>:<controller-
                         port>/rests/data/...) otherwise it  will  just  start  simulated  devices  and  skip the
                         execution of PATCH requests
  --device-count DEVICES-COUNT
                         Number of simulated netconf devices to  spin.  This  is  the number of actual ports open
                         for the devices. (default: 1)
  --devices-per-port DEVICES-PER-PORT
                         Amount of config files  generated  per  port  to  spoof  more  devices than are actually
                         running (default: 1)
  --schemas-dir SCHEMAS-DIR
                         Directory containing yang  schemas  to  describe  simulated  devices.  Some schemas e.g.
                         netconf monitoring and inet types are included by default
  --notification-file NOTIFICATION-FILE
                         Xml  file  containing  notifications  that  should  be  sent  to  clients  after  create
                         subscription is called
  --initial-config-xml-file INITIAL-CONFIG-XML-FILE
                         Xml file containing initial simulatted configuration to be returned via get-config rpc
  --starting-port STARTING-PORT
                         First port for simulated device.  Each  other  device  will  have previous+1 port number
                         (default: 17830)
  --generate-config-connection-timeout GENERATE-CONFIG-CONNECTION-TIMEOUT
                         Timeout to be generated in initial config files (default: 1800000)
  --generate-config-address GENERATE-CONFIG-ADDRESS
                         Address to be placed in generated configs (default: 127.0.0.1)
  --generate-configs-batch-size GENERATE-CONFIGS-BATCH-SIZE
                         Number of connector configs per generated file (default: 1)
  --distribution-folder DISTRO-FOLDER
                         Directory where the karaf distribution for controller is located
  --ssh {true,false}     Whether to use ssh for transport or just pure tcp (default: true)
  --exi {true,false}     Whether to use exi to transport xml content (default: true)
  --debug {true,false}   Whether to use debug log level instead of INFO (default: false)
  --md-sal {true,false}  Whether to use md-sal datastore instead of default simulated datastore. (default: false)
  --time-out TIME-OUT    the maximum time in seconds for executing each PATCH request (default: 20)
  --ip IP                Ip address which will be used for creating  a  socket address.It can either be a machine
                         name, such as java.sun.com, or  a  textual  representation  of its IP address. (default:
                         0.0.0.0)
  --thread-pool-size THREAD-POOL-SIZE
                         The number of threads to keep in  the  pool,  when  creating a device simulator. Even if
                         they are idle. (default: 8)
  --rpc-config RPC-CONFIG
                         Rpc config file. It can be used to  define  custom rpc behavior, or override the default
                         one.Usable for testing buggy device behavior.

Supported operations

Testtool default simple datastore supported operations:

get-schema

returns YANG schemas loaded from user specified directory,

edit-config

always returns OK and stores the XML from the input in a local variable available for get-config and get RPC. Every edit-config replaces the previous data,

commit

always returns OK, but does not actually commit the data,

get-config

returns local XML stored by edit-config,

get

returns local XML stored by edit-config with netconf-state subtree, but also supports filtering.

(un)lock

returns always OK with no lock guarantee

create-subscription

returns always OK and after the operation is triggered, provided NETCONF notifications (if any) are fed to the client. No filtering or stream recognition is supported.

Note: when operation=”delete” is present in the payload for edit-config, it will wipe its local store to simulate the removal of data.

When using the MD-SAL datastore testtool behaves more like normal NETCONF server and is suitable for crud testing. create-subscription is not supported when testtool is running with the MD-SAL datastore.

Notification support

Testtool supports notifications via the --notification-file switch. To trigger the notification feed, create-subscription operation has to be invoked. The XML file provided should look like this example file:

<?xml version='1.0' encoding='UTF-8' standalone='yes'?>
<notifications>

<!-- Notifications are processed in the order they are defined in XML -->

<!-- Notification that is sent only once right after create-subscription is called -->
<notification>
    <!-- Content of each notification entry must contain the entire notification with event time. Event time can be hardcoded, or generated by testtool if XXXX is set as eventtime in this XML -->
    <content><![CDATA[
        <notification xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0">
            <eventTime>2011-01-04T12:30:46</eventTime>
            <random-notification xmlns="http://www.opendaylight.org/netconf/event:1.0">
                <random-content>single no delay</random-content>
            </random-notification>
        </notification>
    ]]></content>
</notification>

<!-- Repeated Notification that is sent 5 times with 2 second delay inbetween -->
<notification>
    <!-- Delay in seconds from previous notification -->
    <delay>2</delay>
    <!-- Number of times this notification should be repeated -->
    <times>5</times>
    <content><![CDATA[
        <notification xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0">
            <eventTime>XXXX</eventTime>
            <random-notification xmlns="http://www.opendaylight.org/netconf/event:1.0">
                <random-content>scheduled 5 times 10 seconds each</random-content>
            </random-notification>
        </notification>
    ]]></content>
</notification>

<!-- Single notification that is sent only once right after the previous notification -->
<notification>
    <delay>2</delay>
    <content><![CDATA[
        <notification xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0">
            <eventTime>XXXX</eventTime>
            <random-notification xmlns="http://www.opendaylight.org/netconf/event:1.0">
                <random-content>single with delay</random-content>
            </random-notification>
        </notification>
    ]]></content>
</notification>

</notifications>

Connecting testtool with controller Karaf distribution

Auto connect to OpenDaylight

You can set up the testtool to automatically connect to the Controller. When you provide the --controller-ip and --controller-port parameters, the testtool will send a PATCH request to the Controller netconf-topology with the device connector configuration.

java -jar netconf-testtool-[VERSION]-executable.jar --device-count 10 --controller-ip 127.0.0.1 --controller-port 8181 --debug true

Running testtool and OpenDaylight on different machines

The testtool binds by default to 0.0.0.0 so it should be accessible from remote machines. However you need to set the parameter “generate-config-address” (when using autoconnect) to the address of machine where testtool will be run so OpenDaylight can connect. The default value is localhost.

Executing operations via RESTCONF on a mounted simulated device

Simulated devices support basic RPCs for editing their config. This part shows how to edit data for simulated device via RESTCONF.

Test YANG schema

The controller and RESTCONF assume that the data that can be manipulated for mounted device is described by a YANG schema. For demonstration, we will define a simple YANG model:

module test {
  yang-version 1;
  namespace "urn:opendaylight:test";
  prefix "tt";
  revision "2014-10-17";

  container cont {
    leaf l {
      type string;
    }
  }
}

Save this schema in file called test@2014-10-17.yang and store it a directory called test-schemas/, e.g., your home folder.

Editing data for simulated device

  • Start OpenDaylight

  • Install odl-netconf-topology and odl-restconf-nb features

  • Start the device with following command:

    java -jar netconf-testtool-[VERSION]-executable.jar --controller-ip 127.0.0.1 --controller-port 8181 --debug true --schemas-dir ~/test-schemas/

  • Check that you can see config data for simulated device by executing GET request to:

    http://localhost:8181/rests/data/network-topology:network-topology/topology=topology-netconf/node=17830-sim-device/yang-ext:mount?content=config

  • The data should be just and empty data container

  • Now execute edit-config request by executing a POST request to:

    http://localhost:8181/rests/data/network-topology:network-topology/topology=topology-netconf/node=17830-sim-device/yang-ext:mount

    with headers:

    Accept application/xml
Content-Type application/xml

    and payload:

    <cont xmlns="urn:opendaylight:test">
  <l>Content</l>
</cont>

  • Response should be 201 with empty body

  • Check that you can see modified config data for simulated device by executing GET request to

    http://localhost:8181/rests/data/network-topology:network-topology/topology=topology-netconf/node=17830-sim-device/yang-ext:mount?content=config

  • Check that you can see the same modified data in operational for simulated device by executing GET request to

    http://localhost:8181/rests/data/network-topology:network-topology/topology=topology-netconf/node=17830-sim-device/yang-ext:mount?content=nonconfig

Warning

Data will be mirrored in operational datastore only when using the default simple datastore.

Testing User defined RPC

The NETCONF test-tool allows using custom RPC. Custom RPC needs to be defined in yang model provide to test-tool along with parameter --schemas-dir.

The input and output of the custom RPC should be provided with --rpc-config parameter as a path to the file containing definition of input and output. The format of the custom RPC file is xml as shown below.

Example YANG model file (stored in folder ~/test-schemas/):

module example-ops {
  namespace "urn:example-ops:reboot";
  prefix "ops";

  import ietf-yang-types {
    prefix "yang";
  }

  revision "2016-07-07" {
    description "Initial version.";
    reference "example document.";
  }

  rpc reboot {
    description "Reboot operation.";
    input {
      leaf delay {
        type uint32;
        units "seconds";
        default 0;
        description "Delay in seconds.";
      }
      leaf message {
        type string;
        description "Log message.";
      }
    }
  }
}

Example payload (RPC config file ~/tmp/customrpc.xml):

<rpcs>
  <rpc>
    <input>
      <reboot xmlns="urn:example-ops:reboot">
        <delay>300</delay>
        <message>message</message>
      </reboot>
    </input>
    <output>
      <rpc-reply xmlns="urn:ietf:params:xml:ns:netconf:base:1.0">
        <ok/>
      </rpc-reply>
    </output>
  </rpc>
</rpcs>

Start the device with following command:

java -jar netconf-testtool-[VERSION]-executable.jar --controller-ip 127.0.0.1 --controller-port 8181 --schemas-dir ~/test-schemas/ --rpc-config ~/tmp/customrpc.xml --debug=true

Example of use:

POST http://localhost:8181/rests/operations/network-topology:network-topology/topology=topology-netconf/node=17830-sim-device/yang-ext:mount/example-ops:reboot

With body:

<?xml version="1.0" encoding="UTF-8" ?>
<input xmlns="urn:example-ops:reboot">
  <delay>300</delay>
  <message>message</message>
</input>

If successful the response should be 204.

Note

A working example of user defined RPC can be found in TestToolTest.java class of the tools[netconf-testtool] project.

Known problems

Slow creation of devices on virtual machines

When testtool seems to take unusually long time to create the devices use this flag when running it:

-Dorg.apache.sshd.registerBouncyCastle=false

Too many files open

When testtool or OpenDaylight starts to fail with TooManyFilesOpen exception, you need to increase the limit of open files in your OS. To find out the limit in linux execute:

ulimit -a

Example sufficient configuration in linux:

core file size          (blocks, -c) 0
data seg size           (kbytes, -d) unlimited
scheduling priority             (-e) 0
file size               (blocks, -f) unlimited
pending signals                 (-i) 63338
max locked memory       (kbytes, -l) 64
max memory size         (kbytes, -m) unlimited
open files                      (-n) 500000
pipe size            (512 bytes, -p) 8
POSIX message queues     (bytes, -q) 819200
real-time priority              (-r) 0
stack size              (kbytes, -s) 8192
cpu time               (seconds, -t) unlimited
max user processes              (-u) 63338
virtual memory          (kbytes, -v) unlimited
file locks                      (-x) unlimited

To set these limits edit file: /etc/security/limits.conf, for example:

*         hard    nofile      500000
*         soft    nofile      500000
root      hard    nofile      500000
root      soft    nofile      500000

“Killed”

The testtool might end unexpectedly with a simple message: “Killed”. This means that the OS killed the tool due to too much memory consumed or too many threads spawned. To find out the reason on linux you can use following command:

dmesg | egrep -i -B100 'killed process'

Also take a look at this file: /proc/sys/kernel/threads-max. It limits the number of threads spawned by a process. Sufficient (but probably much more than enough) value is, e.g., 126676