Monitoring System Hardware
You monitor system hardware in these ways, using:
Retrieve Hardware Information Using decode-syseeprom
decode-syseeprom command enables you to retrieve information about
the switch’s EEPROM. If the EEPROM is writable, you can set values on
cumulus@switch:~$ decode-syseeprom TlvInfo Header: Id String: TlvInfo Version: 1 Total Length: 114 TLV Name Code Len Value -------------------- ---- --- ----- Product Name 0x21 4 4804 Part Number 0x22 14 R0596-F0009-00 Device Version 0x26 1 2 Serial Number 0x23 19 D1012023918PE000012 Manufacture Date 0x25 19 10/09/2013 20:39:02 Base MAC Address 0x24 6 00:E0:EC:25:7B:D0 MAC Addresses 0x2A 2 53 Vendor Name 0x2D 17 Penguin Computing Label Revision 0x27 4 4804 Manufacture Country 0x2C 2 CN CRC-32 0xFE 4 0x96543BC5 (checksum valid)
/usr/cumulus/bin/decode-syseeprom [-a][-r][-s [args]][-t]
|-h||Displays the help message and exits.|
|-a||Prints the base MAC address for switch interfaces.|
|-r||Prints the number of MACs allocated for switch interfaces.|
|-s||Sets the EEPROM content if the EEPROM is writable. args can be supplied in command line in a comma separated list of the form <field>=<value>, …. Illegal characters in field names and values include the comma (,) and equals sign (=). Fields that are not specified default to their current values. If args are supplied in the command line, they will be written without confirmation. If args is empty, the values will be prompted interactively.|
|-j, –json||Displays JSON output.|
|-t TARGET||Prints the target EEPROM (board, psu2, psu1) information. Note: Some systems that use a BMC to manage sensors (such as the Dell Z9264, Facebook Voyager, and Facebook Wedge-100) do not provide the PSU EEPROM contents. This is because the BMC connects to the PSUs via I2C and the main CPU of the switch has no direct access.|
|–serial||Prints the device serial number.|
|-m||Prints the base MAC address for management interfaces.|
|–init||Clears and initializes the board EEPROM cache|
You can also use the
dmidecode command to retrieve hardware
configuration information that’s been populated in the BIOS.
You can use
apt-get to install the
lshw program on the switch, which
also retrieves hardware configuration information.
Monitor System Units Using smond
smond daemon monitors system units like power supply and fan,
updates their corresponding LEDs, and logs the change in the state.
Changes in system unit state are detected via the
smond utilizes these registers to read all sources, which impacts the
health of the system unit, determines the unit’s health, and updates the
smonctl to display sensor information for the various system
cumulus@switch:~$ sudo smonctl Board : OK Fan : OK PSU1 : OK PSU2 : BAD Temp1 (Networking ASIC Die Temp Sensor ): OK Temp10 (Right side of the board ): OK Temp2 (Near the CPU (Right) ): OK Temp3 (Top right corner ): OK Temp4 (Right side of Networking ASIC ): OK Temp5 (Middle of the board ): OK Temp6 (P2020 CPU die sensor ): OK Temp7 (Left side of the board ): OK Temp8 (Left side of the board ): OK Temp9 (Right side of the board ): OK
When the switch is not powered on,
smonctl shows the PSU status as
BAD instead of POWERED OFF or NOT DETECTED. This is a known
On the Dell S4148 switch,
smonctl shows PSU1 and PSU2; however in the
sensors output, both PSUs are listed as PSU1.
Some switch models lack the sensor for reading voltage information, so
this data is not output from the
For example, the Dell S4048 series has this sensor and displays power and voltage information:
cumulus@dell-s4048-ON:~$ sudo smonctl -v -s PSU2 PSU2: OK power:8.5 W (voltages = ['11.98', '11.87'] V currents = ['0.72'] A)
Whereas the Penguin Arctica 3200C (Celestica Seastone) does not:
cumulus@cel-sea:~/tmp$ sudo smonctl -v -s PSU1 PSU1: OK
The following table shows the
smonctl command options.
smonctl [OPTION]... [CHIP]...
|-s SENSOR, –sensor SENSOR||Displays data for the specified sensor.|
|-v, –verbose||Displays detailed hardware sensors data.|
For more information, read
man smond and
Monitor Hardware Health Using sensors
sensors command provides a method for monitoring the health of
your switch hardware, such as power, temperature and fan speeds. This
command executes lm-sensors.
Even though you can use the
sensors command to monitor the health of
your switch hardware, the
smond daemon is the recommended method for
monitoring hardware health. See
Monitor System Units Using smond above.
cumulus@switch:~$ sensors tmp75-i2c-6-48 Adapter: i2c-1-mux (chan_id 0) temp1: +39.0 C (high = +75.0 C, hyst = +25.0 C) tmp75-i2c-6-49 Adapter: i2c-1-mux (chan_id 0) temp1: +35.5 C (high = +75.0 C, hyst = +25.0 C) ltc4215-i2c-7-40 Adapter: i2c-1-mux (chan_id 1) in1: +11.87 V in2: +11.98 V power1: 12.98 W curr1: +1.09 A max6651-i2c-8-48 Adapter: i2c-1-mux (chan_id 2) fan1: 13320 RPM (div = 1) fan2: 13560 RPM
Output from the
sensors command varies depending upon the switch
hardware you use, as each platform ships with a different type and
number of sensors.
The following table shows the
sensors command options.
sensors [OPTION]... [CHIP]...
|-c, –config-file||Specify a config file; use
|-s, –set||Executes set statements in the config file (root only);
|-f, –fahrenheit||Show temperatures in degrees Fahrenheit.|
|-A, –no-adapter||Do not show the adapter for each chip.|
|--bus-list||Generate bus statements for
[CHIP] is not specified in the command, all chip info will be printed.
Example chip names include:
Monitor Switch Hardware Using SNMP
The Net-SNMP documentation is discussed here.
Keep the Switch Alive Using the Hardware Watchdog
Cumulus Linux includes a simplified version of the
daemon from the standard
writes to a file called
/dev/watchdog periodically to keep the switch
from resetting, at least once per minute. Each write delays the reboot
time by another minute. After one minute of inactivity where
wd_keepalive doesn’t write to
/dev/watchdog, the switch resets
itself. The watchdog is enabled by default on all supported switches,
and starts when you boot the switch, before
To enable the hardware watchdog, edit the
/etc/watchdog.d/<your_platform> file and set
run_watchdog to 1:
To disable the watchdog, edit the
run_watchdog to 0:
Then stop the daemon:
cumulus@switch:~$ sudo systemctl stop wd_keepalive.service
You can modify the settings for the watchdog — like the timeout setting
and scheduler priority — in its configuration file,