OBSOLESCENCE WARNING

This implementation for an unattended SQM station is now obsolete and I do not recommend its use.

Unattended Sky Quality Meter Station (V3).

Generalities

About light pollution, SQM, and related projects.

As a amateur astronomer, light pollution is a phenomena which I usually must deal. Over the years, more and more light pollution become a growing problem for the activities of amateur and professional astronomers. Many institutions are involved with this issue:

There is a hardware stuff that is more and more used for measuring the light pollution: The "Sky Quality Meter" from Unihedron. This piece of hardware (basically a photometer) is sold in different varieties, that permit various options for connect it to a computer:

Many organizations around the world centralizes the light pollutions measurements. With the present project I pretend implement a Raspberry Pi based full Unattended light measurement station suitable for server data to some of these organizations, in particular REECL SQM network.

Unattended station concept...

The idea is to have a 24/7 computer with a SQM attached and serve data to some centralized site. Due the price, size and electrical consumption of this computer is a important facet, a Raspberry Pi based station becomes a attractive solution

The Unattended station should be these characteristics:

Pieces of the Puzzle.

For the impatient: downloadable disk system image template

You can implement your own Raspberry Pi unattended SQM data collector by download a image template

The MD5SUM code for this zip file is 7e34fecfc57fa28db0a4510c87747fa0 and for the uncompressed image file is 90e355d499f15ec70bf99833cd8c7427

You can "burn" this image to a SD card in the usually way:

# dd bs=8225280 if=2015-08-25-wheezy-raspbian-PySQM_ready-kernel4.1.6-staticIP192.168.0.251.img of=/dev/mmcblk0; sync

or similar. Please also note that this image is configured with a static IP: 192.168.0.251: you should edit the file /etc/network/interfaces for adapt to your local network scheme.

The default user is the usually one: pi with password raspberry.

By the way, you will need a SD card with at least 4GB in size. Please note that nevertheless the size of the root file system is quite conservative, so don't forget increase it to the maximum of your SD card, for example by using the raspi-config utility after boot your Raspberry Pi.

For more information about how to "writing an image to the SD card", please refer to the official raspberrypi.org site

Once the image was burned into a SD card, you mus edit the files in order to adapt it to you case:

Known issues: Changes you should make in the downloaded image

After upload the previous pre-configured SD image, check following issues that had been discovered. You should fix them after burn the SD from the pre-configured image for avoid problems...

At present time, all known issues are reflected in the system image template.

Setup details: how you can make it

Get account in gmail.com, eu.storagemadeeasy.com and dropbox.com

For the proposed solution, a account in gmail.com, eu.storagemadeeasy.com and dropbox.com should be created. gmail.com is needed for notification stuff and as base for create the account in eu.storagemadeeasy.com and dropbox.com.

eu.storagemadeeasy.com is a cloud storage broker service that have some interesting characteristics:

Loggin in the eu.storagemadeeasy.com page and create a folder named DropBox. Inside, create another folder for store your SQM data. For example, we will use a folder with the name OAF-SQM01 (inside the DropBox folder). Also, add your DropBox account as cloud provider and config the syncronization of the folder DropBox in Storage Made Easy with the DropBox account...

This screenshot show the DropBox account is associate to a subfolder in the Storage Made Easy service:

SME provider

And this one show configuration details for data syncronitation whith DropBox. Please note that use trash for deleted files should be disabled for avoid show some spurious file names in DropBox:

SME configuration

Finally, the data is available in DropBox:

DropBox data

Prepare a system image

We will use Raspbian as Raspberry Pi O.S.. You can download the latest Raspbian image and make you SD by follow this instructions.

Assuming you have a usable Raspbian SD, we begin for upgrade the system. From now, I assume you was become root user in your system. For do this, open a console and type:

$> sudo su

You could remove all the desktop environment stuff and get more disk space by do (credits: RemoveDesktop):

$> apt-get remove xserver* x11-common x11-utils x11-xkb-utils \
x11-xserver-utils xarchiver xauth xkb-data console-setup xinit lightdm \
libx{composite,cb,cursor,damage,dmcp,ext,font,ft,i,inerama,kbfile,klavier,\
mu,pm,randr,render,res,t,xf86}* lxde* lx{input,menu-data,panel,polkit,\
randr,session,session-edit,shortcut,task,terminal} obconf openbox gtk* \
libgtk* scratch tsconf desktop-file-utils

In any case, please remember to resize the file system to maximum amount of space provided by the SD.

Now, the very first job to do is to update/upgrade the system. For this, type:

$> apt-get update

and, after the process ends:

$> apt-get upgrade

and accept the suggestions... only for be sure, you can also type:

$> apt-get dist-upgrade

For clean the downloaded package, do:

$> apt-get --yes autoremove

$> apt-get --yes autoclean

$> apt-get --yes clean

A singularity of the Raspbian O.S. is that (for now) the Linux kernel is packet on the same package that the boot firmware, so the kernel packages for the Raspbian is not used: it use the kernel associated with the boot firmware...

You can update the kernel/firmware with the command:

$> rpi-update

Install and configure mail

A easy (but not so secure) manner for configure a mail system in the Raspberry Pi for use a external account is installing the heirloom-mailx package:

$> apt-get install heirloom-mailx

Edit the file /etc/nail.rc and add at the end some configuration and alias definitions:

#QMS stuff

#Account
set smtp-use-starttls
set ssl-verify=ignore
set smtp=smtp://smtp.gmail.com:587
set smtp-auth=login
set smtp-auth-user=YOURUSER@gmail.com
set smtp-auth-password=YOURPASS
set from="YOURUSER@gmail.com (Friendly Name)"

#Alias
alias infomail \
mymail@gmail.com \
mail1@example.com \
mail2@example.com

alias errormail \
mymail@gmail.com \
mail3@example.com \
mail4@example.com \
mail5@example.com

Execute now some commands due (really weak) security considerations:

$> chown root.mail /etc/nail.rc
$> chown 640 /etc/nail.rc
$> usermod -a -G mail pi

Note that the new group for user pi take effect only since the next login (as user pi).

Install and configure davfs

The necessary package for mount the cloud storage by using the webdav protocol must be installed:

$> apt-get install davfs2

During the installation, you will be prompted about permit webdav mounts to unprivileged user: You should answer "Yes" to this question.

After this, this line must be added this line to the file /etc/fstab:

https://webdaveu.storagemadeEasy.com/DropBox/OAF-SQM01  /home/pi/SQM/data  davfs _netdev,rw,user,noauto,uid=1000,gid=1000,file_mode=644 0 0

Please note that this line assume that the folder OAF-SQM01 was created throw the eu.storagemadeeasy.com web interface. Some other name could be used, at your choice.

Now, as user pi, create the directory /home/pi/.davfs2/ and inside it the file /home/pi/.davfs2/secrets whit this content:

/home/pi/SQM/data       "YOURUSER@gmail.com"        "YOURSMEPASSWORD"

And, for security reasons, do:

$> chmod 600 /home/pi/.davfs2/secrets

An slow and/or transitory network issues can produce some error that could be minimized by edit the file /home/pi/.davfs2/davfs2.conf and set the option use_locks to 0 (use_locks 0). Credits about this.

Also, PySQM suffers a very poor input/output performance, which produce a bottleneck that results in a high consumption of bandwidth when cloud storage is used. For avoid this issue, add to the end of the file /home/pi/.davfs2/davfs2.conf the lines:

use_locks 0
cache_size 128

[/home/pi/SQM/data]
## saving bandwidth:
## 10 minutes delay in file and dir refresh...
file_refresh    600
dir_refresh     600
## 10 minutes delay upload. Hope this save bandwidth
delay_upload    600

Please note that in this example, use_locks and cache_size are general options for all possible davfs2 mounted filesystem, but the other ones only affect to the filesystem mounted under /home/pi/SQM/data.

Installation and configuration of PySQM

Lets go with the core of the system: the PySQM software, developed by the Jaime Zamorano team and that can be found at its page.

From the project web page: "PySQM the UCM open source software to read, plot and store data from SQM photometers" Nievas Rosillo, Miguel and Zamorano Calvo, Jaime (2014) UCM eprint (PDF).

Installing necessary dependences.

Assuming you are using a SQM-LU model, install the necessary dependences by running, as root these commands:

$> apt-get install python-pip

$> apt-get install python-matplotlib

$> apt-get install ttf-bitstream-vera

$> apt-get install python-dev

$> pip install pyephem

Please note that some other dependences, as python-serial and python-numpy should be yet installed in the default Raspbian distribution. Also note that the default Raspbian distribution have installed various python versions, but the default one is the 2.7: that is the required version by PySQM.

There is a little change about matplotlib stuff that you should to in order to use the library in a not interactive way: Edit the file /etc/matplotlibrc and change the line:

backend      : TkAgg

by this one:

backend      : Agg

A little hack for avoid certain aesthetics issues due a bug in one of the matplotlib related packages (about the fonts path used).

$> cd /usr/share/fonts/truetype

$> ln -s lyx ttf-lyx

Also, lets to get permission to read the port associate to the SQM-LU to the user pi. This is done by the command:

$> usermod -a -G dialout

Note that the new group for user pi take effect only since the next login (as user pi).

Download and configure PySQM

As user pi, in its home, make a new directory named SQM and go into it:

$> cd; mkdir SQM; cd SQM

All the follow stuff in this section will be made into this /home/pi/SQM. Lets to download the PySQM software from PYSQM. It can be done directly by the command:

$> wget http://guaix.fis.ucm.es/sites/default/files/luminica_files/PySQM.tar.gz

and uncompress it:

$> tar -xvzf PySQM.tar.gz

some other directories should be create for future work:

$> mkdir data logs

In our working directory (/home/pi/SQM) should be appears a file named config.py. Edit it. As a primer step, this line should be added immediately after the very first one... So the two very first lines of this file should be:

#!/usr/bin/env python
# -*- coding: utf-8 -*-

This not avoid the programmer's advice: "Non-ASCII characters are not supported in the config.py file". But, at least, permit you to use utf-8 characters into the coments... :D

In the rest of the this file (config.py) you must configure your particular configuration, observatory name, position, etc...

In _device_addr you should put _device_addr = '/dev/ttyUSB0' if you are using a SQM-LU. By the way, we are using UT as system time, so its related variable should be _computer_timezone = 0. About the variable _local_timezone you should use the timezone for where your system is configured. For UTC, it should be: _local_timezone = 0

About the path for data store, it should point to the same directory where you mount the webdav cloud storage. This is: monthly_data_directory = "/home/pi/SQM/data" You can maintain the defaults for the others path...

Main script

Script /home/pi/SQM/data/unattended_sqm.sh:

As user pi, you must create the following script:

#!/bin/bash

. /etc/profile.d/sqm_environment.sh

PYSQMPATH="$(dirname $0)"

CLOUDSTORAGEPATH="${PYSQMPATH}/data"

LOGFILE="${PYSQMPATH}/logs/unattended_sqm.log"

PYSQMOUT="${PYSQMPATH}/logs/pysqm.stdout"

PYSQMERR="${PYSQMPATH}/logs/pysqm.stderr"

MOUNTCHECK="mountpoint -q ${CLOUDSTORAGEPATH}"

PYSQMCOMMAND="python -m pysqm"

## Use timeout command to avoid conection issues
MOUNT_TIMEOUT=90

MOUNT_PID_FILE="/var/run/mount.davfs/$(echo ${CLOUDSTORAGEPATH} |sed 's/^\///' |sed  's/\/$//' |sed 's/\//-/g').pid"

VALID_PROCESS_STATUS=("S" "R")

_MSG="$(date -R -u): ${SQM_NAME}"

function killpysqm {
   for i in $(pgrep -f -x "${PYSQMCOMMAND}"); do
      kill -TERM $i; wait
      # just in case:
      kill -KILL $i
   done
}

function execpysqm {
   cd ${PYSQMPATH}
   nohup ${PYSQMCOMMAND} >>${PYSQMOUT} 2>>${PYSQMERR} < /dev/null &
   MSG="======> ${_MSG} --> STARTING PROCESS"
   echo ${MSG} >>${PYSQMOUT}
   echo ${MSG} >>${PYSQMERR}
   MSG="PID = $(pgrep -f -x "${PYSQMCOMMAND}")"
   echo ${MSG} >>${PYSQMOUT}
   echo ${MSG} >>${PYSQMERR}
}

function checkpysqm {
   # first, check for bad state process and kill them
   for i in $(pgrep -f -x "${PYSQMCOMMAND}"); do
      _STATUS=$(ps h o state p $i)
      _VALID="NO"
      for _ITEM in ${VALID_PROCESS_STATUS[*]}; do
           [[ "${_STATUS}" == "${_ITEM}" ]] && _VALID="YES"
      done
      if [ "${_VALID}" == "$NO" ];then kill -KILL $i; fi
   done; wait

   NUMOFPYSQM=$(pgrep -c -f -x "${PYSQMCOMMAND}"); wait

   if [ "${NUMOFPYSQM}" == "0" ]; then
      execpysqm
      MSG="${_MSG} --> Starting: \"${PYSQMCOMMAND}\"."
      echo "$MSG" >>$LOGFILE
      echo "$MSG" |mail -v -s "${_MSG}: Starting" infomail
   elif [ "${NUMOFPYSQM}" == "1" ]; then
      MSG="${_MSG} --> \"${PYSQMCOMMAND}\" already running."
      echo "$MSG" >>$LOGFILE
   else
      killpysqm
      MSG="${_MSG} --> Terminating: too many \"${PYSQMCOMMAND}\" processes."
      echo "$MSG" >>$LOGFILE
      echo "$MSG" |mail -v -s "${_MSG}: Terminating" errormail
   fi
}

### BEGIN

${MOUNTCHECK}
if [ $? -ne 0 ]; then
    # if exist a pid file, delete it:
    rm -f ${MOUNT_PID_FILE}
    timeout ${MOUNT_TIMEOUT} mount "${CLOUDSTORAGEPATH}"
    ${MOUNTCHECK}
    if [ $? -ne 0 ]; then
        MSG="${_MSG} --> ${CLOUDSTORAGEPATH} mount failed."
        echo "$MSG" >>$LOGFILE
        echo "$MSG" |mail -v -s "${_MSG}: Mount Failed" errormail
        killpysqm
    else
        MSG="${_MSG} --> ${CLOUDSTORAGEPATH} mount successfully."
        echo "$MSG" >>$LOGFILE
        echo "$MSG" |mail -v -s "${_MSG}: Mount successfully" infomail
        checkpysqm
    fi
else
    MSG="${_MSG} --> ${CLOUDSTORAGEPATH} already mounted."
    echo "$MSG" >>$LOGFILE
    checkpysqm
fi

sync

This file must be made executable by using the command:

$> chmod u+x unattended_sqm.sh

Script /root/bin/check_ip.sh:

As user root, you must create first the directory:

$> mkdir /root/bin

and then create the following script (/root/bin/check_ip.sh) inside:

#!/bin/bash

. /etc/profile.d/sqm_environment.sh

LOGFILE=${SQMDIR}/logs/check_ip.log

_DATE=$(date -R -u)
## Use timeout command to avoid conection issues
_TIMEOUT=3
# An array for list external ip service discovers to probe
_IPSERVICES=('http://ipecho.net/plain' 'http://ipv4.icanhazip.com' 'http://ifconfig.me/ip')

for _USEDIPSERV in ${_IPSERVICES[*]}; do
   NOWIP=$(timeout ${_TIMEOUT} wget -qO- ${_USEDIPSERV}); wait
   if [ ! -z "$NOWIP" ]; then break; fi
done
if [ -z "$NOWIP" ]; then
   NOWIP=unset
   _USEDIPSERV=none
fi

if [ "${NOWIP}" == "${EXTERNAL_IP}" ]; then
   echo "${_DATE} from ip discover: ${_USEDIPSERV} --> External IP unchanged: ${EXTERNAL_IP}" >>${LOGFILE}
   sync

else
   echo "This file contains the last detected external IP" >${LASTIPFILE}
   echo "Dont edit by hand" >>${LASTIPFILE}
   echo "${_DATE} from ip discover: ${_USEDIPSERV}" >>${LASTIPFILE}
   echo >>${LASTIPFILE}
   echo "${NOWIP}" >>${LASTIPFILE}
   sync

   echo "${_DATE} from ip discover: ${_USEDIPSERV} --> New external IP detected: ${NOWIP}" >>${LOGFILE}
   sync
   echo "${_DATE}: RaspberryPi ${SQM_NAME} --> New external IP (discover: ${_USEDIPSERV}) detected: ${NOWIP}"| mail -v -s "RasPi ${SQM_NAME} new external IP - ${_DATE}" errormail

fi

exit 0

This file must be made executable by using the command:

$> chmod u+x check_ip.sh

Using cron for do the system (reasonable) fault tolerant

For a (reasonable) fault tolerant mechanism, lets use cron for periodically run the above scripts. As i before said, it check if all is right, and re-run the stuff if it is necessary.

As root, the files /etc/cron.d/unattended_sqm and /etc/cron.d/check_ip must be created with the following content:

File /etc/cron.d/unattended_sqm

SHELL=/bin/sh
PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin

## SQM: unattended_sqm
## Every 10 minutes
*/10 *   * * *  pi      /home/pi/SQM/unattended_sqm.sh

File /etc/cron.d/check_ip

SHELL=/bin/sh
PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin

## SQM: check_ip
## Every 6 hours
0    */6 * * *    root      /root/bin/check_ip.sh

Once created, restart cron with the command:

$> /etc/init.d/cron restart

Log subsystem

The main scripts generates log files. These files could grow indefinitely, so the logrotate utility can be used for avoid this stuff. For this, create, as root two files named /etc/logrotate.d/unattended_sqm and /etc/logrotate.d/check_ip with the contents:

File /etc/logrotate.d/unattended_sqm

/home/pi/SQM/logs/unattended_sqm.log {
       daily
       rotate 8
       delaycompress
       compress
       noolddir
       missingok
}
/home/pi/SQM/logs/pysqm.stderr /home/pi/SQM/logs/pysqm.stdout {
       weekly
       rotate 4
       delaycompress
       copytruncate
       compress
       noolddir
       missingok
}

File /etc/logrotate.d/cheick_ip

/home/pi/SQM/logs/check_ip.log {
       weekly
       rotate 4
       delaycompress
       compress
       noolddir
       missingok
}

Tasks on start up and halt

At this point, we must do that some tasks become to automatically be made at system start up and halt. These tasks should be:

Dealing with the clock at boot time

The Raspberry Pi have not a real internal clock (RTC), so when it is powered of the time recorded in the last shutdown is re-established Nevertheless in a few seconds the ntpd stuff do its work and the system become in time, the use of the recorded time at shutdown at the very early boot time can be a crap. But if we can a network available, the system cat be request a more accurate time at network interface up. This time is not accurate, but is very better that use the recorded time at shutdown. For do this stuff, simply install the ntpdata package by using the command:

$> apt-get install ntpdate

Defining environment variables

Create the file /etc/profile.d/sqm_environment.sh with the following content:

## SQM environment

export LASTIPFILE=/tmp/last_ip.txt

export SQMDIR=/home/pi/SQM/

export SQM_NAME="OAF-SQM01"

if [ -f ${LASTIPFILE} ]; then
   EXTERNAL_IP=$(grep -o '[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}' ${LASTIPFILE} |tail -1)
fi
if [ -z "$EXTERNAL_IP" ]; then EXTERNAL_IP=unset; fi
export EXTERNAL_IP

The rc.local file

Edit the file /etc/rc.local. some extra content must be added at the end of the file. The total content of the file will be as follow:

#!/bin/sh -e
#
# rc.local
#
# This script is executed at the end of each multiuser runlevel.
# Make sure that the script will "exit 0" on success or any other
# value on error.
#
# In order to enable or disable this script just change the execution
# bits.
#
# By default this script does nothing.

# Print the IP address
_IP=$(hostname -I) || true
if [ "$_IP" ]; then
  printf "My IP address is %s\n" "$_IP"
fi

## Begin of SQM stuff:
/root/bin/check_ip.sh
. /etc/profile.d/sqm_environment.sh

_DATE=$(date -R -u)

# The -v (verbose) flag ensure send entire message before the command return
su -c "echo \"${_DATE}: RaspberryPi ${SQM_NAME} data collector powered on with external IP: ${EXTERNAL_IP}\"| mail -v -s \"RasPi ${SQM_NAME} POWERED ON - ${_DATE}\" errormail; wait" pi

su -c /home/pi/SQM/unattended_sqm.sh pi

wait
## End of SQM stuff

exit 0

Ensure that this file have permissions -rwxr-xr-x and root as owner.

The rc.local.shutdown file

A new file /etc/rc.local.shutdown must be created with the content:

#!/bin/sh -e
#
# rc.local
#
# This script is executed at the end of each multiuser runlevel.
# Make sure that the script will "exit 0" on success or any other
# value on error.
#
# In order to enable or disable this script just change the execution
# bits.
#
# By default this script does nothing.

## Begin of SQM stuff:

. /etc/profile.d/sqm_environment.sh

_DATE=$(date -R -u)

# The -v (verbose) flag ensure send entire message before the command return
su -c "echo \"${_DATE}: RaspberryPi ${SQM_NAME} data collector shutdown.\"| mail -v -s \"RasPi ${SQM_NAME} SHUTDOWN - ${_DATE}\" errormail; wait" pi

wait
## End of SQM stuff:

exit 0

Ensure that this file have permissions -rwxr-xr-x and root as owner.

Modifying and active the rc.local service

The actual rc.local service must be modified for the new scheme. Edit the file /etc/init.d/rc.local for become content this stuff:

#! /bin/sh
### BEGIN INIT INFO
# Provides:          rc.local
# Required-Start:    $all
# Required-Stop:     $all
# Default-Start:     2 3 4 5
# Default-Stop:      0 1 6
# Short-Description: Run /etc/rc.local and /etc/rc.local.shutdown if it exist
### END INIT INFO


PATH=/sbin:/usr/sbin:/bin:/usr/bin

. /lib/init/vars.sh
. /lib/lsb/init-functions

do_start() {
        if [ -x /etc/rc.local ]; then
                [ "$VERBOSE" != no ] && log_begin_msg "Running local boot scripts (/etc/rc.local)"
                /etc/rc.local
                ES=$?
                [ "$VERBOSE" != no ] && log_end_msg $ES
                return $ES
        fi
}

do_stop() {
        if [ -x /etc/rc.local.shutdown ]; then
                [ "$VERBOSE" != no ] && log_begin_msg "Running local boot scripts (/etc/rc.local.shutdown)"
                /etc/rc.local.shutdown
                ES=$?
                [ "$VERBOSE" != no ] && log_end_msg $ES
                return $ES
        fi
}

case "$1" in
    start)
        do_start
        ;;
    restart|reload|force-reload)
        echo "Error: argument '$1' not supported" >&2
        exit 3
        ;;
    stop)
        do_stop
        ;;
    *)
        echo "Usage: $0 start|stop" >&2
        exit 3
        ;;
esac

After this, for the change do effect, we must do:

$> /usr/sbin/update-rc.d -f rc.local remove
$> /usr/sbin/update-rc.d rc.local defaults

Wake up a dummy service for external monitoring (optional)

The system is reasonably robust but, what about the possibility of a network failure?. In this case, is not possible for the system to notify the failure from inside the local network. For be possible this notify of network failure you can enable a dummy service and monitoring it from outside the local network using, for example, a service as uptimerobot.com.

The best option is wake up a very light web server and serve a dummy page. My personal preference as light web server is webfs. For installing it:

$> apt-get install webfs

After this, create a directory for store the dummy web page and put the appropriate permissions/owner:

$> cd /srv
$> mkdir www
$> chown www-data. www

Inside this new directory (/srv/www), put a index.html file with a dummy content (for example, information about the system and URLs of shared data)

Fix the adequate permissions/owner for this file:

$> chmod 644 index.html; chown www-data. index.html

For last, you must fix some configuration stuff in /etc/webfsd.conf file. The lines I changed for my case were:

web_root="/srv/www"
web_port="1180"
web_index="index.html"

After these changes, restart the webfs service:

$> /etc/init.d/webfs restart

Once a dummy web page is available, you can create an account in uptimerobot.com for monitoring and report issues.

Please note: in this example dummy web service I am using the port number 1180. You can use whatever you like, but in order to become the service available from outside your local network, you should configure your router for redirect the incoming packages to port 1180 to the local IP which the SQM data collector station is configured.

Perhaps, you also should do a similar redirection if you want to use the SSH' access service from outside your local network. As a suggestion, you can change thesshdport and thepi` user password...

Physical deployment of the SQM and begin of operations

The built prototype was installed in the "Observatorio Astronómino de Forcarei (OAF)" (Astronomical Observatory of Forcarei).

In the next image, we can see the look of the installed prototype (after take the picture, the support was updated using metalic clamps). The Raspi is inside the building. Note that only three wires are connected to the Raspi (power, network, USB) and only one of them (the USB wire that connec with the SQM) must cross the wall.

The SQM station

The data :D

The data can be browsed by share the DropBox (final cloud storage) data directory. As example for our case you can show the data in this URL:

https://www.dropbox.com/sh/spze9pleyvg2kjs/AADtv1UDcFBCTtUOrH9Vr0h7a?lst

It is also possible get the wole data file in zip format throw the URL:

https://www.dropbox.com/sh/spze9pleyvg2kjs/AADtv1UDcFBCTtUOrH9Vr0h7a?dl=1

And also, from you account in DropBox you can share the URL of the generated graphic. In my case this image can be acceded throw the URL:

https://www.dropbox.com/s/rw5lvwb80b18r5o/SQM01_Observatorio_de_Forcarei.png?raw=1

By using this URL, the image can be embedded in any web page, as for example:

http://rdlazaro.info/sqm/

Chagelog since system deployment

Before May 2015

Time configuration parameters.

In the first release, the time related parameter were (due my country is in the +1 timezone):

_computer_timezone = +0
_local_timezone = +1

But, if you configure your Raspi for use UTC time (as i do), these parameters should be:

_computer_timezone = 0
_local_timezone = 0

According to the PySQM main author, the parameter _computer_timezone is not used in recents versions (is a legacy parameter) and _local_timezone controls how the time axis is labeled in the graphics (hours un UTC, UTC+1 and so on).

Change of cloud storage provider and use a broker.

In a early stage PowerFolder service was used as primary cloud storage, but file access from davfs2 was disabled by PowerFolder staff. Strange behavior was detected when accessing the files generated by PySQM in the file system mounted with the WebDAV protocol using davfs2.

If you editing (eg using nano or vim) or you read (using eg cat or less) them, their contens was corrupted, and the data record and graphics generation fails.

But it you don't "touch" the files from the Raspi, the stuff works right.

This behavior is produced, according PowerFolder staff, because some html access method related to the webdav protocol were disables by security reasons at PowerFolder server side.

PowerFolder staff claims that them expect add the davfs2 client to a "white list" that should permit a complete functionality when webdav filesystem is mounted from Linux... in 2 or 3 months (since May 2015).

Due these issues, the primary cloud storage system was changed to Storage Made Easy

Currently, PowerFolder staff claims the issue is resolved, but anyway, PowerFolder service is not yet used.

May 2015

Join the `REECL SQM Network' .

At May 2015, the Unattended Sky Quality Meter Station mounted at Forcarei obsevatory becomes join the REECL SQM Network.

As commented before, the PowerFolder service produced some problems. Looking for an alternative, we found the Storage Made Easy cloud storage service. This storage service can be mounted by using the webdav protocol as described before in this document. But this storage service have NOT the other desired characteristics (basically, the use of directs links for share contents). However, Storage Made Easy have a surprising capability that compensates this deficiency: some other cloud storage (as for example, DropBox) can be associate to Storage Made Easy accound and synchronize the stored files whith the associate external cloud storage.

So the Storage Made Easy service was used for mount in the Raspi the remote data storege, and a DropBox account was also associate with the Storage Made Easy contents. So the data can be acceded throw a DropBox account.

After these changes, we fulfill all requirements for join the REECL SQM Network.

August 2015

Bandwidth consumption issue.

PySQM is a nice software, but it was not designed with performance in mind. It suffers a very poor input/output performance, which produce a bottleneck that results in a high consumption of bandwidth when cloud storage is used. For avoid this issue, you must tunning the davfs2' stuff by add the file/home/pi/.davfs2/davfs2.conf` these lines:

[/home/pi/SQM/data]
## saving bandwidth:
## 10 minutes delay in file and dir refresh...
file_refresh    600
dir_refresh     600
## 10 minutes delay upload. Hope this save bandwidth
delay_upload    600

December 2015

Some system tuning was made. I hope some of them avoid the issue about remote (davfs2) filesystem "frozen"...

Logrotate (and other stuff) runs now during the day.

In an usual system, logrotate runs during the night, but due the nature of the SQM measurements, this was changed it to running during the day. We hope that this avoid a possible system overload in the middle of the SQM data recording...

Really, ALL processes scheduled at night were rescheduled at day.

For do it, chage the file /etc/crontab from its original (only relevant lines showed):

17 *    * * *   root    cd / && run-parts --report /etc/cron.hourly
25 6    * * *   root    test -x /usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.daily )
47 6    * * 7   root    test -x /usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.weekly )
52 6    1 * *   root    test -x /usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.monthly )

to:

17 *    * * *   root    cd / && run-parts --report /etc/cron.hourly
25 12   * * *   root    test -x /usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.daily )
47 12   * * 7   root    test -x /usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.weekly )
52 12   1 * *   root    test -x /usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.monthly )

Use a RAM based file system as cache for davfs2:

Create the mount point /home/pi/.davfs2/cache_tmpfs (as user pi) and add this line to /etc/fstab (edit it as user root):

tmpfs   /home/pi/.davfs2/cache_tmpfs    tmpfs  rw,size=384M,nr_inodes=5k,noexec,nodev,nosuid,uid=1000,gid=1000,mode=1700   0  0

Mount it and change the cache path at /home/pi/.davfs2/davfs2.conf (cache_dir) parameter. You must also remount the remote filesystem.

Increase the cache size and number of entries for files.

These are fixed by the parameters cache_size and table_size in the /home/pi/.davfs2/davfs2.conf file.

After all these changes, the /home/pi/.davfs2/davfs2.conf looks as:

use_locks 0
cache_size 256
table_size 4096
cache_dir ~/.davfs2/cache_tmpfs

[/home/pi/SQM/data]
file_refresh    600
dir_refresh     600
delay_upload    600

Contact

If you have any comment or correction, please feel free for send me a e-mail about anything that you consider interesting at monje314-2005 at yahoo dot es.

Cheers!!!