24 July, 2017

WinTec GRays 2 GPS device feeding to NTP

USB driver

When I got a wintec g-rays2, it didn't work on USB out of the box with whatever laptop I had (probably a macbook), but was OK on Bluetooth, so that's what I stuck with.

Now, eight years later, I plugged it into one of my Raspberry Pis and it appears as /dev/ttyUSB1 by magic!

Getting NMEA sentences in minicom

minicom --device=/dev/ttyUSB1 --baud=4800 gives some garbled stuff every second, so I'm receiving the NMEA sentences but at the wrong serial port settings.

The manual didn't give any help but a bit of fiddling reveals sensible looking output at 57600 baud, 8N1 (here you can see where I live).

$GPRMC,121815.000,A,5130.3697,N,00003.7216,W,0.00,148.43,240717,,,A*72     
$GPGGA,121815.000,5130.3697,N,00003.7216,W,1,05,2.9,46.2,M,47.0,M,,0000*70 
$GPGSA,A,3,21,26,31,27,16,,,,,,,,4.3,2.9,3.2*38                            
$GPGSV,3,1,12,05,03,021,28,10,08,157,19,21,67,086,34,26,69,175,39*7C       
$GPGSV,3,2,12,29,12,083,,07,08,333,,31,06,193,26,20,27,059,24*72           
$GPGSV,3,3,12,49,,,35,27,43,274,31,16,72,283,39,18,26,132,23*4D 

These lines are spewed out once a second without needing to send any start command to the GPS unit.

Getting ntpd to pay attention

$GPRMC and $GPGGA are the relevant sentences for ntpd, according to the manual.

I already have ntpd runnning on this Pi, with an MSF receiver configured already.

This gives a /dev/gps0 (at least until reboot):

cd /dev
sudo ln -s ttyUSB1 gps0

and this line in /etc/ntpd.conf makes ntpd look for NMEA time sentences on /dev/gps0:

server 127.127.20.0 mode 67

The mode, decimal 67, means hexadecimal 0x43: 0x01 listen for GPRMC, 0x02 listen for GPGGA, 0x40 use 57600 baud.

And after a restart, tada! (although apparently a 160ms delay compared to all my other time sources. ick)

pi@faeroe /dev $ ntpq --peers
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 2001:8b0:1638:9 81.2.122.172     2 u   39   64   17    0.952  157.794   0.783
 ntp2.aa.net.uk  195.66.241.2     2 u   41   64   17   14.923  162.505   0.671
 SHM(2)          .MSF.            0 l    -   64    0    0.000    0.000   0.000
*GPS_NMEA(0)     .GPS.            0 l   55   64    7    0.000   -5.660   0.819
 tyne.cqx.ltd.uk 81.2.122.172     2 u   46   64   17    0.674  157.858   0.759

03 July, 2017

/etc/hosts - the gift that keeps on giving

/etc/hosts

The gift that keeps on giving, when you're paid on an hourly rate to do tech support. It is like the regexps of DNS: "I know, I'll modify /etc/hosts" ... now you have two problems.

  • You modify /etc/hosts because you want to override the global DNS view of name N.
  • Things works!
  • Because things work, you don't undo your change in /etc/hosts. After all, things work.
  • A year passes.
  • You forget that you made the change.
  • Suddenly on your machine only, but no one else's, you can't access the website at N any more.
  • Hours of debugging!

Why modify /etc/hosts? Testing a new version of a site under real name (only on your own machines though), or because DNS is broken and you don't know why, or because you didn't plan your DNS change so now the caches are not updating fast enough and you are impatient.

/etc/hosts doesn't have an "eventual consistency" mehanism in place - DNS caches eventually expire and at least give vaguely consistent behaviour across the whole internet over a long enough period of time. /etc/hosts will never converge.