Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
731 lines (593 sloc) 28.1 KB

Emulators

OpenDCRE comes with pre-built emulators for each supported devicebus type so that it can emulate API commands and functionality in the absence of supported hardware. This is especially useful for testing and experimenting with OpenDCRE.

PLC Emulator

The PLC Emulator runs as a background process in the OpenDCRE container itself. It uses socat to create a connected pair of virtual TTY devices that can be used to simulate serial communications without hardware. The main difference between the emulated PLC comms and hardware PLC comms is that the emulator does not require the additional steps of configuring the serial device and initializing the SIG60 PLC modem.

The emulator itself is a primitive Python process that is provided a configuration file on startup to map board/device ids to raw packet readings. These packet readings are returned to the device on the other end of the virtual serial connection. Faults, repeated values, cycling values, or no value returns are supported by the PLC emulator through its config file. State can also be preserved in cases where an incoming command mutates state (e.g. turns on an LED) - state honoring may be enabled/disabled in the emulator configuration for a board/device.

Configuration Example

{
  "boards": [
    {
      "board_id": "00000001",
      "firmware_version": "OpenDCRE Emulator v1.0.0 - Standalone Server",
      "devices": [
        {
          "host_info": {
            "repeatable": true,
            "responses": [
              "i10.10.1.16,htest-server0"
            ]
          },
          "has_state": true,
          "boot_target": "B0",
          "pxe" : "B1",
          "no_override": "B2",
          "hdd": "B0",
          "asset_info": {
            "repeatable": true,
            "responses": [
              "Quanta,0001,Winterfell,S1234567,rack mount chassis,P1234567,S1234567,A1234567,Quanta,P1234567,Winterfell,S1234567,v1.2.0"
            ]
          },
          "device_type": "system",
          "device_id": "0001"
        },
        {
          "read": {
            "repeatable": true,
            "responses": [
              4100, 4100, 4000, 4000, 3900, 3900, 3800, 3800, 3700, 3700,
              3800, 3800, 3900, 3900, 4000, 4000, 4100, 4100, 4200, 4200
            ]
          },
          "write": {
            "repeatable": true,
            "responses": [
              "W1"
            ]
          },
          "device_type": "fan_speed",
          "device_id": "0002"
        },
        {
          "read": {
            "repeatable": true,
            "responses": [
              4100, 4100, 4000, 4000, 3900, 3900, 3800, 3800, 3700, 3700,
              3800, 3800, 3900, 3900, 4000, 4000, 4100, 4100, 4200, 4200
            ]
          },
          "write": {
            "repeatable": true,
            "responses": [
              "W1"
            ]
          },
          "device_type": "fan_speed",
          "device_id": "0003"
        },
        {
          "device_id": "0004",
          "device_type": "power",
          "has_state": true,
          "on": [
            "0,10000,0,0", "0,11000,0,0", "0,12000,0,0", "0,13000,0,0",
            "0,14000,0,0", "0,15000,0,0", "0,14000,0,0", "0,13000,0,0",
            "0,12000,0,0", "0,11000,0,0"
          ],
          "off": "64,0,0,0",
          "power": [
            "0,10000,0,0", "0,11000,0,0", "0,12000,0,0", "0,13000,0,0",
            "0,14000,0,0", "0,15000,0,0", "0,14000,0,0", "0,13000,0,0",
            "0,12000,0,0", "0,11000,0,0"
          ]
        },
        {
          "device_id": "0005",
          "device_type": "led",
          "has_state": true,
          "read": 0,
          "write": 0,
          "on": 1,
          "off": 0
        },
        {
          "read": {
            "repeatable": true,
            "responses": [
              28.78, 29.77, 30.75, 31.84, 32.82, 33.81, 34.89, 35.88, 36.96, 37.94,
              38.93, 40.21, 41.27, 42.33, 43.39, 44.45, 45.61, 46.57, 47.63, 48.69,
              49.75, 48.69, 47.63, 46.57, 45.61, 44.45, 43,39, 42.33, 41.27, 40.21,
              38.93, 37.94, 36.96, 35.88, 34.89, 33.81, 32.82, 31.84, 30.75, 29.77
            ]
          },
          "device_type": "temperature",
          "device_id": "2000"
        },
        {
          "read": {
            "repeatable": true,
            "responses": [
              28.78, 29.77, 30.75, 31.84, 32.82, 33.81, 34.89, 35.88, 36.96, 37.94,
              38.93, 40.21, 41.27, 42.33, 43.39, 44.45, 45.61, 46.57, 47.63, 48.69,
              49.75, 48.69, 47.63, 46.57, 45.61, 44.45, 43,39, 42.33, 41.27, 40.21,
              38.93, 37.94, 36.96, 35.88, 34.89, 33.81, 32.82, 31.84, 30.75, 29.77
            ]
          },
          "device_type": "temperature",
          "device_id": "4000"
        }
      ]
    }
  ]
}

Configuration Fields

boards:

A list of boards configured for the emulator, where a single board configuration would represent a single hardware board that should exist.

board_id:

The internal id for a single board. Board Ids in OpenDCRE have a 4 byte width and should be expressed in the config as a 4-byte hex string. Board Ids should be unique across all device instances. For PLC, the board id range starts at 0x00000000.

firmware_version:

The version string for the board, which would be returned by the OpenDCRE “version” command.

devices:

A list of all device configurations which are associated with that given board.

repeatable:

A flag which denotes that the given responses should repeat. This means that when the emulator has cycled through all of the responses in the responses list, it will return to the beginning of the list. If this is set to false, upon reaching the end of the responses list, the emulator will not return data, causing an error to be raised in OpenDCRE (as the emulator will not respond).

responses:

A list of canned responses for the emulator to return.

host_info:

The response(s) to return on a system "host info" command.

asset_info:

The response(s) to return on a system "asset info" command.

read:

The response(s) to return on a "read" command.

write:

The response(s) to return on a "write" command.

on:

The response(s) to return on a "power on" command.

off:

The response(s) to return on a "power off" command.

power:

The response(s) to return on a "power status" command.

has_state:

If true, then state information is preserved relative to the command (e.g. "on" or "off" for power), in which case subsequent reads retrieve a response relative to the persisted state. When state is undefined, the default response (e.g. "power") for the command is returned. When has_state is false, the response relative to an incoming command is returned (e.g. the response for "on" for a power "on" command).

boot_target:

Indicates the response sent back for a "get" of boot target (B0 == no_override, B1 == pxe, B2 == hdd).

pxe:

The response sent when boot target of PXE is set.

no_override:

The response sent when boot target of no_override is set.

hdd:

The response sent when boot target of HDD is set.

device_type:

Indicates the type of device that a device entry represents. This is also the device_type reported back in OpenDCRE REST API scan results. Valid device types include:

  • thermistor
  • power
  • humidity
  • pressure
  • led
  • system
  • fan_speed
  • temperature
.. versionchanged:: 1.2
    In previous releases, a device type of ``none`` indicated that no device is present at a given ``device_id`` on the
    given board, and may be ignored.  In OpenDCRE v1.2 the ``none`` device type has been removed.

device_id:

The internal device id of the device being configured, expressed as a 2-byte numeric value as a hex string. In most cases, a device id of "0001" is sufficient.

Additional Information on Configuration Fields

Device Type

A field corresponding to the action supported for a given device type is required. A map of device types to supported actions is below:

Device Type Action Supported
thermistor read
temperature read
power power
humidity read
pressure read
led read, write
fan_speed read, write
system asset info, boot target

Read

For the read action's field in the OpenDCRE emulator configuration, two fields may be configured relating to the responses returned from a read command for the given device.

First, the repeatable field may be set to true or false, depending on whether it is desirable for the list of responses set in the responses field to repeat in a round-robin fashion, or if a device should stop returning data after its response list has been exhausted.

The responses field is a list of zero or more values that may be returned for a given read command. The raw values are converted (where necessary) by the built-in OpenDCRE conversion functions, based on the given device_type.

When a list of values is provided for responses, the emulator iterates sequentially through the items in that list, until the list is exhausted (if repeatable is set to "true", then the emulator returns to the beginning of the list).

An empty responses list means the device returns no data, which translates to a 500 error for the read command at the OpenDCRE REST API level (useful for simulating errors). To always return the same single value, a responses list with a single element, and repeatable set to "true" will suffice.

Read Response Format

The table below describes the response format for each device type for read commands to the emulator.

Device Type Format
thermistor integer, converted by OpenDCRE
temperature numeric, sent back as numeric value (e.g. 28.78)
humidity numeric, converted by OpenDCRE
led integer, 1 is on and 0 is off; all other values are errors
fan_speed integer, sent back as integer value (e.g. 4100)

Values that do not conform to the above formats will result in errors to read requests made to the emulator, as they would on the device bus.

Write

For the write action's field in the OpenDCRE emulator configuration, two fields may be configured, relating to the responses returned from a write command for the given device. The fields are laid out and function in the same manner as read fields.

Write Response Format

The table below describes the response format for each device type for write commands to the emulator.

Device Type Format
led string - W1 is successful, while W0 is unsuccessful; all other values are errors.
fan_speed string - W1 is successful, while W0 is unsuccessful; all other values are errors.

Values that do not conform to the above formats will result in errors to write requests made to the emulator, as they would on the device bus.

Writing to a device from OpenDCRE to the emulator does not currently result in any state change for a corresponding device in the emulator. That functionality may be added in a future release.

Power

For the power action's field in the OpenDCRE emulator configuration, similar fields are present - repeatable and responses.

For every power command (e.g. on/off/cycle/status) issued to a power device in the OpenDCRE emulator, a response is returned from the responses list, which may be repeatable or non-repeatable. The values in the responses list correspond to power status values returned over PMBUS from the hot swap controller on an OCP server, and are expressed as an integer value in the emulator configuration (see example above). OpenDCRE converts the raw response to a friendly power status result using its built-in conversion functions.

Other Notes

The OpenDCRE emulator is also used for testing purposes, and additional emulator configurations may be found under the /opendcre/opendcre_southbound/tests/data directory of the OpenDCRE Docker container.

An invalid emulator configuration will cause the OpenDCRE emulator to fail to start or function properly.

Additional features of the emulator that may be used by advanced users or hardware/protocol developers include:

  • Ability to send back raw bytes for responses to scan, version, read, write, and power commands. In tests, this can be seen where a list (or list of lists) of integer values is specified for a given response. Special sentinel values (999, 10xx) are used to place sequence numbers and checksums into the packet stream.
  • Ability to support command retries in cases of invalid packets, line noise, etc.
  • Ability to support 'scan-all' command and retries using time-division multiplexing; success and failure scenarios may be implemented for various configurations. See the test-scanall tests.

Running the Emulator

To run the PLC emulator, simply specify the startup script for the PLC emulator.

docker run -p 5000:5000 vaporio/opendcre ./start_opendcre_plc_emulator.sh

or, if using docker-compose:

opendcre:
  image: vaporio/opendcre
  command: ./start_opendcre_plc_emulator.sh
  ports:
    - 5000:5000

The examples above will start the emulator with the default configuration file, found at /opendcre/opendcre_southbound/emulator/plc/data/example.json. To specify different emulator configurations, simply pass that file as an argument to the emulator start script. Note that if the non-default emulator configuration is not built into the OpenDCRE image, it will need to be volume-mounted in, e.g.

docker run \
    -p 5000:5000 \
    -v `pwd`/emulator_config:/opendcre/new_emulator_config.json \
    vaporio/opendcre \
    ./start_opendcre_plc_emulator.sh /opendcre/new_emulator_config.json

IPMI Emulator

For IPMI communications, there is an IPMI emulator which exists as a Dockerized Python multithreaded UDP server that accepts inbound UDP IPMI packets, processes them, and returns a response based on the emulator configuration and internal state.

The IPMI Emulator, which is perhaps better described as a BMC Emulator, is stateful where applicable. For example, one can set the boot target or LED state on the emulator and a subsequent examination of either should reveal the state to be the new values it was set to.

The IPMI Emulator is primarily designed to work with pyghmi, as that is the library used within OpenDCRE to issue IPMI commands. To accommodate pyghmi, the emulator supports: - HMAC-SHA1-96 integrity checking - RAKP_HMAC_SHA1 authentication - AES_CBC_128 encryption. The encrypted mode can be tested using pyghmi or ipmitool.

For ease of use and simplicity for debugging, it also supports no authentication/encryption which allows all bytes to be examined (e.g. with Wireshark). This unencrypted mode can be tested using ipmitool.

The IPMI emulator is largely just a framing device which unpacks incoming requests and packs outgoing responses. The actual logic to handle commands is often simple, typically just returning values either from internal state or emulator configuration.

Configuration Example

The IPMI Emulator configuration lives in the opendcre/opendcre_southbound/emulator/ipmi/data directory and is built into the emulator’s Docker image (at the same path, starting at root). Configurations can be changed by either - Modifying the source configurations and rebuilding the Docker image - Mounting in configuration overrides with Docker volumes.

There are four configuration files associated with the IPMI emulator

Note

All of the raw bytes specified in these config files were taken off the wire (using Wireshark) when communicating with a real BMC.

bmc.json

bmc.json contains the configurations for the mock BMC that is the IPMI emulator. It allows the specification of device info, chassis info, channel authentication capabilities, and dcmi configurations. Generally, the configurations specified in this file are the raw bytes that make up the IPMI responses.

{
  "device": {
    "device_id": "20",
    "device_revision": "01",
    "device_availability": "03",
    "minor_firmware_revision": "16",
    "ipmi_version": "02",
    "additional_device_support": "bf",
    "manufacturer_id": 47488,
    "product_id": 2566
  },
  "chassis": {
    "current_power_state": "01",
    "last_power_event": "00",
    "misc_state": "40",
    "bootdev": "no_override"
  },
  "channel_auth_capabilities": {
    "channel": "01",
    "version_compatibility": "96",
    "user_capabilities": "06",
    "supported_connections": "03",
    "oem_id": 21317,
    "oem_auxiliary_data": "00"
  },
  "capabilities": {
    "hpm": ["81", "b4", "cb", "20", "08", "3e", "c1", "d9"],
    "picmg": ["81", "b4", "cb", "20", "10", "00", "c1", "0f"],
    "vita": ["81", "b4", "cb", "20", "14", "00", "c1", "0b"]
  },
  "dcmi": {
    "power": {
      "current_watts": [185, 188, 186, 189, 188, 192, 195, 199, 210, 211, 213, 211, 212],
      "min_watts": 150,
      "max_watts": 250,
      "avg_watts": 200,
      "reporting_interval_ms": 305000
    },
    "capabilities": {
      "1": ["dc", "01", "05", "02", "00", "01", "07"],
      "2": ["dc", "01", "05", "02", "00", "00", "00", "00", "00"],
      "3": ["dc", "01", "05", "02", "20", "00"],
      "4": ["dc", "01", "05", "02", "ff", "ff", "ff"],
      "5": ["dc", "01", "05", "02", "01", "00"]
    }
  }
}

fru.json

fru.json contains the raw configuration data for the mock BMC’s FRU. The config file specifies the FRU inventory area and the raw data that makes up the FRU.

{
  "inventory_area": 1024,
  "device_access": 0,
  "data": [
    "01", "00", "00", "01", "06", "00", "00", "f8", "01", "05", "00",
    "00", "00", "00", "ca", "53", "75", "70", "65", "72", "6d", "69",
    "63", "72", "6f", "c0", "ca", "20", "20", "20", "20", "20", "20",
    "20", "20", "20", "20", "c0", "c0", "c1", "00", "00", "00", "00",
    "00", "00", "fc", "00", "01", "03", "00", "c0", "c0", "c0", "c0",
    "ca", "20", "20", "20", "20", "20", "20", "20", "20", "20", "20",
    "c0", "c0", "c1", "00", "00", "b1"
  ]
}

sdr.json

sdr.json contains the raw configuration data for the mock BMC’s SDR. This includes the version, record count, free space, latest addition, latest erase, and operation support. The configuration for the actual SDR records is not specified in this file, but in sdr_entries.json.

{
  "sdr_version": 1.5,
  "record_count": 21,
  "free_space": 1663,
  "latest_addition_ts": 0,
  "latest_erase_ts": 0,
  "operation_support": "2f"
}

sdr_entries.json

sdr_entries.json is the config file where all device records belonging to the SDR are defined. The number of devices defined in this config should match the device count specified in sdr.json. For each record, an id, sensor type, data, readings, event messages, and threshold comparison field should be specified. The sensor type field is not used by the IPMI emulator, but is used as a convenient means of labeling the record with a human-readable description.

{
  "records": [
    {
      "id": "0000",
      "sensor_type": "System Temp",
      "data": [
        "04", "00", "51", "01", "36", "20", "00", "11", "07", "01", "7d", "68",
        "01", "01", "80", "7a", "80", "7a", "3f", "3f", "80", "01", "00", "00",
        "01", "00", "00", "00", "00", "00", "07", "2d", "4a", "fc", "7f", "80",
        "4f", "4d", "4b", "f7", "f9", "fb", "02", "02", "00", "00", "00", "cb",
        "53", "79", "73", "74", "65", "6d", "20", "54", "65", "6d", "70"
      ],
      "readings": [
        49, 49, 48, 47, 48
      ],
      "event_messages": "c0",
      "threshold_comparison": ["c0"]
    },
    {
      "id": "0047",
      "sensor_type": "CPU Temp",
      "data": [
        "47", "00", "51", "01", "33", "20", "00", "12", "03", "01", "7f", "68",
        "01", "01", "80", "7a", "80", "7a", "3f", "3f", "80", "01", "00", "00",
        "01", "00", "00", "00", "00", "00", "07", "1e", "59", "fc", "7f", "80",
        "5f", "5a", "55", "f5", "f8", "fb", "02", "02", "00", "00", "00", "c8",
        "43", "50", "55", "20", "54", "65", "6d", "70"
      ],
      "readings": [
        41, 40, 41, 41
      ],
      "event_messages": "c0",
      "threshold_comparison": ["c0"]
    },
    {
      "id": "008a",
      "sensor_type": "CPU FAN",
      "data": [
        "8a", "00", "51", "01", "32", "20", "00", "41", "1d", "01", "7d", "68",
        "04", "01", "95", "7a", "95", "7a", "3f", "3f", "00", "12", "00", "00",
        "b9", "00", "00", "c0", "00", "01", "07", "80", "aa", "14", "ff", "00",
        "b2", "af", "ac", "10", "11", "12", "01", "01", "00", "00", "00", "c7",
        "43", "50", "55", "20", "46", "41", "4e"
      ],
      "readings": [
        34, 34, 35, 34, 33
      ],
      "event_messages": "c0",
      "threshold_comparison": ["c0"]
    },
    ...
    (abridged for brevity)
  ]
}

Getting the Emulator

Since the IPMI Emulator is a standalone image, it needs to either be pulled from DockerHub, or built from Dockerfile.

From DockerHub,

docker pull vaporio/ipmi-emulator-x64

From Dockerfile, first navigate to opendcre_southbound/emulator/ipmi. Then, you can build the IPMI emulator image with

make build-x64

Running / Using the Emulator

Running the emulator in isolation is straightforward enough. Once you have the image, you can run it either with docker:

docker run --name ipmi-emulator -p 623:623/udp vaporio/ipmi-emulator-x64

or with docker-compose

docker-compose -f ipmi-emulator.yml up --build -d ipmi-emulator

where ipmi-emulator.yml contains

ipmi-emulator:
  container_name: ipmi-emulator
  image: vaporio/ipmi-emulator-x64
  command: ./start_ipmi_emulator.sh
  ports:
    - 623:623/udp

While other emulators (e.g. the PLC emulator) are built in to OpenDCRE and can be run from the same container, the IPMI emulator must be run from a separate container, as shown above.

This is done in part for emulator isolation, but also because it allows for more flexible test setups. For instance, with the emulator running in a separate container it is possible to spin up multiple emulator instances, each with their own configuration, to emulate OpenDCRE performance against different BMC models. Additionally, with docker-compose, OpenDCRE can have multiple proxies to the same emulator to simulate a full rack, cluster, or multi-cluster of BMCs.

Warning

When using the IPMI emulator in a proxied fashion, where multiple composefile links point to the same emulator, the number of requests issued against the emulator can become an issue, especially under high network latency, where the requests back up and time out.

When run locally, this had caused the emulator to freeze up and communications between OpenDCRE and the emulator fail. One solution to this is to run the IPMI Emulator on a separate instance/machine when there will be heavy load placed upon it. This will ensure that it is given enough machine resources to operate at full capacity - although network latency can then become an issue.

Running the emulator on a separate instance/machine is recommended even without heavy load, for stability and performance.

Above, we describe how to run an IPMI emulator. Some additional configuration will need to happen with OpenDCRE in order for it to register the IPMI emulator as a usable interface.

The networking between the emulator and OpenDCRE is determined by the BMC config used by OpenDCRE. For example, if there were a BMC config containing the record:

{
  "bmc_ip": "localhost",
  "username": "ADMIN",
  "password": "ADMIN"
}

we would want the emulator to be running on the same machine as OpenDCRE, as localhost should resolve to the emulator.

The containers can also be linked in the composefile, if running on the same machine, so we can reference the emulator using the container name as a hostname:

{
  "bmc_ip": "ipmi-emulator",
  "username": "ADMIN",
  "password": "ADMIN"
}

Of course, a plain IP for the machine running the emulator can be supplied as the bmc_ip without any need to create container links.

An example (abridged) composefile with the two containers linked is as follows:

opendcre:
  image: vaporio/opendcre-core-x64
  command: ./start_opendcre.sh
  ports:
    - 5000:5000
  links:
    - ipmi-emulator

ipmi-emulator:
  image: vaporio/ipmi-emulator-x64
  ports:
    - 623:623/udp

Note that here, the OpenDCRE instance was started without running any other emulator. While it is possible (and fine) to run the IPMI emulator alongside any of the serial emulators, keeping things isolated to IPMI-only for testing is usually prudent.

Redfish Emulator

Warning

Redfish support is in beta as of OpenDCRE v1.3.0

Like the IPMI emulator, the Redfish Emulator is a standalone Dockerized python application. It runs a simple Flask webserver that serves up statically defined configuration data. It supports the basic Redfish commands and is stateful, for operations where state can be preserved (e.g. turning an LED on). By default, the Redfish emulator runs on port 5040. This can be changed by updating the Dockerfile and specifying the correct port mapping at run time.

Configuration

The configuration files which make up the Redfish emulator backend are too numerous to include here - instead, see the Redfish mockups which the configuration hierarchy is based off of.

To re-configure the Redfish emulator, either a new emulator image can be built with the new configuration placed in the emulator's Resources directory, or it can be volume mounted in over the emulator's Resources directory.

Getting the Emulator

Since the Redfish Emulator is a standalone image, it needs to either be pulled from DockerHub, or built from Dockerfile.

From DockerHub,

docker pull vaporio/redfish-emulator-x64

From Dockerfile, first navigate to opendcre_southbound/emulator/redfish. Then, you can build the Redfish emulator image with

make build-x64

Running the Emulator

Running the Redfish emulator is simple, given that the desired configurations (whether they be the default or custom built-in/volume-mounted configurations) are correctly placed in the image:

docker run -p 5040:5040 vaporio/redfish-emulator-x64