Re: [gpsd-users] Control Gpsd

[Alexander Carver]

No, the watchdog in the Pi does not monitor a process ID.  It uses a
kernel module to ping the hardware watchdog timer built into the
processor on a regular basis.  If the kernel stops, the hardware timer
will count down to zero and then force the processor to reset.

If you have another type of watchdog running then perhaps that is how it
works but the watchdog I was talking about is built into the Pi hardware.

On 2016-03-15 11:42, bruno suarez wrote:
> no, but watchdog verifies that the process whose number is in the PID
> file still exists,
>  so it amounts to the same as the script that you gave, right?
> 
> Bruno
> Le 15/03/2016 19:09, Alexander Carver a écrit :
>> There's no guarantee that the PID file will go away if gpsd stops.
>>
>> On 2016-03-15 10:59, bruno suarez wrote:
>>> Another solution without programming is to use watchdog and monitoring
>>> the pid file of gpsd
>>>
>>> This works for me, if gpsd stops the system reboot
>>>
>>> Bruno
>>>
>>> Le 15/03/2016 18:01, bruno suarez a écrit :
>>>> Thanks for the help.
>>>> This is a good start for me ...
>>>> I'll try this, I think in the script (cron) I will restart Pi directly
>>>> because if there is a problem it may be the receiver or other items
>>>>
>>>> Bruno
>>>>
>>>> Le 15/03/2016 17:01, Alexander Carver a écrit :
>>>>> I can't really give you an example because there's many ways of doing
>>>>> it.  It's all up to the programming languages you are most familiar
>>>>> with
>>>>> (and those that run on the Raspberry Pi).  It's also not exactly a one
>>>>> or two line script although there are some very rough ways of doing it
>>>>> in a small script, they're not very robust.
>>>>>
>>>>> But in general you need to write a small program that connects to
>>>>> gpsd,
>>>>> sends the WATCH command so that data streams from the attached GPS
>>>>> receiver and then detect if that data stops flowing (because gpsd
>>>>> crashed) or the data isn't updating (because the GPS receiver
>>>>> failed or
>>>>> disconnected).  In that case you send a system call to kill the PID of
>>>>> the gpsd process then start a new copy.
>>>>>
>>>>>
>>>>> The small script version would be something like (this is completely
>>>>> untested, don't use it)
>>>>>
>>>>> #!/bin/bash
>>>>> GPSDPID=$(ps ax | grep gpsd | grep -v grep | awk '{print $1}')
>>>>> GPSDSTATE=$(ps ax | grep gpsd | grep -v grep | awk '{print $3}'
>>>>> GPSDCMD="/path/to/gpsd --with-parameters"
>>>>> if [ "$GPSDSTATE" == 'Z' ] || [ -z "$GPSDPID" ];
>>>>> then
>>>>>      echo 'gpsd is a zombie or it's completely missing, kill it'
>>>>>      if [ -n "$GPSDPID" ];
>>>>>      then
>>>>>          kill -KILL $GPSDPID
>>>>>          sleep 1
>>>>>      fi
>>>>>
>>>>>      $GPSDCMD
>>>>> fi
>>>>>
>>>>> Basically that script would look at the process list to see if gpsd
>>>>> was
>>>>> there, pull out the PID and the state and then see if the state is a Z
>>>>> (zombie process) or an empty string (the -z) in which case it died and
>>>>> needs to be restarted.  This is a bad script because it can't catch
>>>>> many
>>>>> problems that might leave gpsd running but not really functioning.
>>>>>
>>>>> You are better off using something else to connect directly to gpsd's
>>>>> TCP port and taking action if the data isn't there, the port closes or
>>>>> some other detail you don't like.
>>>>>
>>>>>
>>>>> On 2016-03-14 23:51, bruno suarez wrote:
>>>>>> Hello,
>>>>>>
>>>>>> Thank you for the reply and the idea of chemical heat packs.
>>>>>> The watchdog is expected.
>>>>>>
>>>>>> On monitoring GPSD, ok for the principle, but a sample script would
>>>>>> help
>>>>>> me (I am not a man of script!).
>>>>>>
>>>>>>
>>>>>> Thank you for your answers and tips
>>>>>> cordially
>>>>>> Bruno
>>>>>>
>>>>>> Le 15/03/2016 07:36, Alexander Carver a écrit :
>>>>>>> On 2016-03-14 12:06, bruno suarez wrote:
>>>>>>>> Hello,
>>>>>>>> For a stratospheric balloon project, I usea raspberry Rasbian
>>>>>>>> Jessie and
>>>>>>>> a USB GPS receiver.
>>>>>>>>
>>>>>>>> Everything is OK with gpsd and my GPS receiver.
>>>>>>>> When I plug everything works correctly, gpsd start on the correct
>>>>>>>> port
>>>>>>>> /dev/ttyUSB0.
>>>>>>>>
>>>>>>>> However, I want to make sure that everything works properly during
>>>>>>>> 2/3
>>>>>>>> hours to extremely cold temperature
>>>>>>>>
>>>>>>>> What is the easiest way to automatically control the operation of
>>>>>>>> GPSD
>>>>>>>> and relaunch it(on the same port detected ) if it is stopped or if
>>>>>>>> it is
>>>>>>>> in error.
>>>>>>>>
>>>>>>>> Thank you for your answers and tips
>>>>>>>>
>>>>>>>> Cordially
>>>>>>>> Bruno
>>>>>>> First, pack your payload with heaters.  You can use chemical heat
>>>>>>> packs
>>>>>>> from a pharmacy that are normally used for heating muscles and
>>>>>>> joints
>>>>>>> (these are the plastic bags that you smash to mix the chemicals and
>>>>>>> they
>>>>>>> get warm.)  In an insulated container you should stay within your
>>>>>>> operating limits of the Pi (the lower limit is approximately -20
>>>>>>> Celsius).
>>>>>>>
>>>>>>> After that, just set up a script that watches for data coming from
>>>>>>> gpsd
>>>>>>> and, if there is no data, issue a kill for the PID of gpsd and
>>>>>>> restart
>>>>>>> it.  Since gpsd can support more than one connection you can get
>>>>>>> data
>>>>>>> for your recording/telemetry program and have a completely separate
>>>>>>> program running to watch for a loss of communication.
>>>>>>>
>>>>>>> Also make sure to enable the watchdog timer in the Pi so that it
>>>>>>> will
>>>>>>> reboot if the whole system fails.  Make sure gpsd will start at
>>>>>>> boot as
>>>>>>> well as your data logging/telemetry program and the monitoring
>>>>>>> script.
>>>>>>>
>>>>>>
>>>>
>>>
>>>
>>
> 
> 
>