I check in on my Raspberry Pi 3 running dump1090 and PiAware 3.1
Running smooth as ExLax but my stats show I haven’t checked in for six hours. FFS.
Under ‘anomalies’ it says that 3.5 is available and I should upgrade. OK.
I try the Send Command to Device “Upgrade and Restart PiAware”. In the log window is says “Sorry the log appears to be empty”
Before anyone asks, I can ping flightaware.com. I restarted dump1090 in interactive mode. Planes out the ass. Restarted it in the background. Stats still shows I haven’t checked in for six hours. I looked for a notice of maintenance or known outage. Didn’t see anything.
We started doing automatic upgrades to 3.5 around that time, for those systems that had auto upgrades enabled.
I guess you are in the 5% of systems where the upgrade failed for one reason or another, and it failed in such a way that piaware stopped.
I looked at the logs that did get sent back to FA, and the last we heard from your Pi, it looked liked the package upgrade was fine and piaware was restarting itself; but we never heard back from piaware after that.
OK, so what I saw was that piaware was not running. I did ‘netstat -pn’ and looked for a TCP connection to 70.42.6.224:1200 but saw nothing. I tried to run it from the command line. Got “buffer overflow detected”. Ran it under strace and saw that it is trying to create an IPv6 socket and being told off by the kernel. I did a “modprobe ipv6” then restarted piaware and it’s looking good now…
To recap (for anyone troubleshooting)
ps afx | grep piaware
netstat -pn | grep piaware
lsmod | grep ipv6
modprobe ipv6
service piaware restart or could just ‘nohup piaware &’ and call it a day…
I was afraid I was going to lose my 410 day streak
Hmm, this sounds vaguely familiar but I thought it had been fixed (or at least worked around). But yeah - enabling ipv6 support will fix it.
(You might want to put “ipv6” in /etc/modules so it is loaded on reboot)
edit: ah, right, here it is: ads-b-flight-tracking-f21/problem-with-piaware-3-1-0-t38122.html
It’s a bug somewhere in tcl itself that I haven’t been able to track down yet, triggered by having an ipv6 address for localhost but not having ipv6 socket support.
You can work around it either by enabling ipv6 (as you did) or by removing the ipv6 address from /etc/hosts
I was expecting to do the 3.5 upgrade by hand. Surprise, something good did happen during those meetings I was in earlier today!
But rest assured, Murphy got in a small reminder he’s still around – my automagic backups go off Thursday mornings at 2AM, so a normal backup won’t capture all this new stuff until next week.
But that’s why I have manual backup triggers as well.
