How to set up ‘live’ logging with pywws

Introduction

There are two quite different modes of operation with pywws. Traditionally pywws.hourly would be run at regular intervals (usually an hour) from cron. This is suitable for fairly static websites, but more frequent updates can be useful for sites such as Weather Underground (http://www.wunderground.com/). The newer pywws.livelog program runs continuously and can upload data every 48 seconds.

Getting started

First of all, you need to install pywws and make sure it can get data from your weather station. See How to get started with pywws for details.

If you have previously been using pywws.hourly then disable your ‘cron’ job (or whatever else you use to run it) so it no longer runs. You should never run pywws.hourly and pywws.livelog at the same time.

Try running pywws.livelog from the command line, with a high level of verbosity so you can see what’s happening. Use the pywws-livelog command to run pywws.livelog:

pywws-livelog -vvv ~/weather/data

Within five minutes (assuming you have set a 5 minute logging interval) you should see a ‘live_data new ptr’ message, followed by fetching any new data from the weather station and processing it. Let pywws.livelog run for a minute or two longer, then kill the process by typing ‘<Ctrl>C’.

Changed in version 14.04.dev1194: the pywws-livelog command replaced scripts/pywws-livelog.py.

Configuring file locations

Open your weather.ini file with a text editor. You should have a [paths] section similar to the following (where xxx is your user name):

[paths]
work = /tmp/pywws
templates = /home/xxx/weather/templates/
graph_templates = /home/xxx/weather/graph_templates/
modules = /home/xxx/weather/modules/

Edit these to suit your installation and preferences. work is a temporary directory used to store intermediate files. If your computer uses solid state storage, such as a Raspberry Pi’s SD card, it’s a good idea to make this a “RAM disk” to reduce storage “wear”. templates is the directory where you keep your text template files, graph_templates is the directory where you keep your graph template files, and modules is a directory for any extra modules you write. Don’t use the pywws example directories for your templates, as they will get over-written when you upgrade pywws.

Copy your text and graph templates to the appropriate directories. You may find some of the examples provided with pywws useful to get started. The pywws-version -v command should show you where the examples are on your computer.

New in version 14.04.dev1194: the pywws-version command.

Configuring periodic tasks

In weather.ini you should have a [live] section similar to the following:

[live]
services = []
text = []
plot = []

This section specifies what pywws should do every time it gets a new reading from the weather station, i.e. every 48 seconds. The services entry is a list of online weather services to upload data to, e.g. 'underground' or ('ftp', '24hrs.txt'). The plot and text entries are lists of template files for plots and text files to be processed. You should probably leave all of these blank except for services.

If you use YoWindow (http://yowindow.com/) you can add an entry to the [live] section to specify your YoWindow file, e.g.:

[live]
services = ['underground']
text = ['yowindow.xml']
plot = []

If you don’t already have them, create four more sections in your weather.ini file: [logged], [hourly], [12 hourly] and [daily]. These sections should have similar entries to the [live] section, and specify what to do every time data is logged (5 to 30 minutes, depending on your logging interval), every hour, twice daily and once per day. Add the names of your template files to the appropriate entries, for example:

[logged]
services = ['underground', 'metoffice']
plot = []
text = []

[hourly]
services = [('twitter', 'tweet.txt'),
            ('ftp', '7days.png', '24hrs.png', 'rose_24hrs.png',
                    '24hrs.txt', '6hrs.txt', '7days.txt')]
plot = ['7days.png.xml', '24hrs.png.xml', 'rose_24hrs.png.xml']
text = ['tweet.txt', '24hrs.txt', '6hrs.txt', '7days.txt']

[12 hourly]
services = []
plot = []
text = []

[daily]
services = [('twitter', 'forecast.txt'), ('ftp', '28days.png', 'allmonths.txt')]
plot = ['28days.png.xml']
text = ['forecast.txt', 'allmonths.txt']

Note that the twitter and ftp “services” use files generated by the plot and text items. It’s probably best not to add all of these at once. You could start by uploading one file to your web site, then when that’s working add the remaining web site files. You can add Twitter and other services later on.

New in version 14.05.dev1211: [cron name] sections. If you need more flexibility in when tasks are done you can use [cron name] sections. See weather.ini - configuration file format for more detail.

Create a dedicated user (optional)

As pywws will be running continuously, and contacting various computers on the internet, there is a very remote risk that one of its dependencies has a security flaw that might allow someone to gain unauthorised to your computer. Running pywws as a user with minimal privileges adds a little extra protection.

You can create a user with the adduser command:

sudo adduser --system --disabled-login --shell=/bin/false pywws

The exact syntax may vary according to your operating system. The important thing is to create a user that can’t login, and can’t run sudo, but does have a home directory.

Run in the background

In order to have pywws.livelog carry on running after you finish using your computer it needs to be run as a “background job”. On most Linux / UNIX systems you can do this by putting an ampersand (‘&’) at the end of the command line. Running a job in the background like this doesn’t always work as expected: the job may suspend when you log out. It’s much better to run as a proper UNIX ‘daemon’ process.

Using systemd

On recent versions of Linux the systemd service manager makes it easy to create a daemon process. The service is defined in a file /etc/systemd/system/pywws.service:

[Unit]
Description=pywws weather station live logging
After=time-sync.target

[Service]
Type=simple
User=pywws
Restart=on-failure
ExecStart=/usr/local/bin/pywws-livelog -v -l systemd /home/pywws/data/

The [Unit] section says pywws shouldn’t start until the computer has set its clock correctly. This is important on computers without a battery-backed real time clock, such as the Raspberry Pi. The [Service] section specifies which user should run pywws and gives the command to run it. The -l systemd option sends log messages to systemd instead of using a normal pywws log file. You can use sudo service pywws start to test the pywws.service file. After starting sudo service pywws status shows if it’s running OK, and the last few log messages:

[email protected]:~ $ sudo service pywws status
● pywws.service - pywws weather station live logging
   Loaded: loaded (/etc/systemd/system/pywws.service; static; vendor preset: enabled)
   Active: active (running) since Thu 2018-08-23 17:49:01 BST; 12min ago
 Main PID: 30946 (pywws-livelog)
   CGroup: /system.slice/pywws.service
           └─30946 /usr/bin/python3 /usr/local/bin/pywws-livelog -v -l systemd /home/pywws/data/

Aug 23 17:49:44 gordon pywws-livelog[30946]: pywws.service.wetterarchivde:server response "{'version': '6.0', 'status': 'SUCCESS'}"
Aug 23 17:49:44 gordon pywws-livelog[30946]: pywws.service.metoffice:OK
Aug 23 17:49:45 gordon pywws-livelog[30946]: pywws.service.openweathermap:OK
Aug 23 17:49:46 gordon pywws-livelog[30946]: pywws.service.cwop:OK
Aug 23 17:49:46 gordon pywws-livelog[30946]: pywws.service.underground:server response "success"
Aug 23 17:57:41 gordon pywws-livelog[30946]: pywws.weatherstation:setting sensor clock 5.33264
Aug 23 17:57:41 gordon pywws-livelog[30946]: pywws.weatherstation:sensor clock drift 1.4672 1.08387
Aug 23 18:00:25 gordon pywws-livelog[30946]: pywws.service.mastodon:OK
Aug 23 18:00:26 gordon pywws-livelog[30946]: pywws.service.twitter:OK
Aug 23 18:00:26 gordon pywws-livelog[30946]: pywws.service.sftp:OK
[email protected]:~ $

If you’d prefer to use a normal pywws log file the pywws.service file might look like this:

[Unit]
Description=pywws weather station live logging
After=time-sync.target

[Service]
Type=simple
User=pywws
Restart=on-failure
PermissionsStartOnly=true
ExecStartPre=/bin/mkdir -p /var/log/pywws
ExecStartPre=/bin/chown -R pywws:nogroup /var/log/pywws/
ExecStart=/usr/local/bin/pywws-livelog -v -l /var/log/pywws/pywws.log /home/pywws/data/

In this example the log file is /var/log/pywws/pywws.log. The directory /var/log/pywws/ might not exist after a reboot (e.g. if /var/log/ has been moved to RAM disk to reduce SD card wear) so ExecStartPre is used to create it and transfer its ownership to the pywws user. PermissionsStartOnly=true ensures the ExecStartPre commands are run as root.

The udev system can be used to start the pywws service when the computer boots or the weather station is plugged into the USB port. (It also stops pywws if the weather station is unplugged.) Create a file /etc/udev/rules.d/39-weather-station.rules as follows:

SUBSYSTEM=="usb" \
, ATTRS{idVendor}=="1941" \
, ATTRS{idProduct}=="8021" \
, OWNER="pywws" \
, TAG+="systemd" \
, ENV{SYSTEMD_WANTS}="pywws.service"

This sets the owner of the weather station’s USB port to pywws, then adds pywws.service to the things systemd should be running.

Using pywws-livelog-daemon

If you can’t use systemd for some reason then the pywws.livelogdaemon program can be used to run pywws as a daemon, if you have the python-daemon library installed:

pywws-livelog-daemon -v ~/weather/data ~/weather/data/pywws.log start

(Note that the log file is a required parameter, not an option.) Unfortunately the python-daemon package appears not to be maintained, and I’ve had problems with it on some Linux versions. You’ll also need to setup something to start pywws automatically.

There are various ways of configuring a Linux system to start a program when the machine boots up. Typically these involve putting a file in /etc/init.d/. A slightly harder problem is ensuring a program restarts if it crashes. My solution to both problems is to run the following script from cron, several times an hour.

#!/bin/sh

export PATH=$PATH:/usr/local/bin

# exit if NTP hasn't set computer clock
[ `ntpdc -c sysinfo | awk '/stratum:/ {print $2}'` -ge 10 ] && exit

pidfile=/var/run/pywws.pid
datadir=/home/jim/weather/data
logfile=$datadir/live_logger.log

# exit if process is running
[ -f $pidfile ] && kill -0 `cat $pidfile` && exit

# email last few lines of the logfile to see why it died
if [ -f $logfile ]; then
  log=/tmp/log-weather
  tail -40 $logfile >$log
  /home/jim/scripts/email-log.sh $log "weather log"
  rm $log
  fi

# restart process
pywws-livelog-daemon -v -p $pidfile $datadir $logfile start

The process id of the daemon is stored in pidfile. If the process is running, the script does nothing. If the process has crashed, it emails the last 40 lines of the log file to me (using a script that creates a message and passes it to sendmail) and then restarts pywws.livelogdaemon. You’ll need to edit this quite a lot to suit your file locations and so on, but it gives some idea of what to do.


Comments or questions? Please subscribe to the pywws mailing list http://groups.google.com/group/pywws and let us know.