Arrgh! Network keeps randomly dropping when I run MX

[mzambo001]

This is driving me nuts. MX running on a Raspberry Pi. Works great. For a while... Then randomly the network goes out. I can ping it but can not ssh into it. It quits updating weather underground and my web site. Could be 30 minutes to a day and a half when it drops.

I don't see anything weird in the syslog or in the MX log. I even tried to set it up on a different Pi. It does the same thing. I'm running 3.0.0 b3028 on a Raspberry Pi 2Model B and moved it to a Raspberry Pi 1Model B. Running rasbian wheezy 2015-05-05.

If anyone has any ideas, please help.

[water01]

Power supply not 2amp rated so it goes into sleep mode and shuts down certain peripherals including network card?

[rogerthn]

Power supply not 2amp rated so it goes into sleep mode and shuts down certain peripherals including network card?

AND
Wired or wireless network?
There are "some" to read regarding Raspberry WiFi drops out after ...

[mzambo001]

Yep, I just verified that the power supply is a 2A unit. The network is the wired connection. That gives me an idea though, maybe I'll put it on wireless next time it goes South. Shouldn't be a long wait...

[rogerthn]

If two different PIs has the same problem I would check elsewhere and if you have GOOD wireless access, try WiFi.
But, wired is normally always better than wireless and maybe you could start with

Code: Select all

sudo apt-get install ethtool
ethtool eth0

and check lines like Speed: and Duplex:

[mzambo001]

One would think that I would have thought of that... I will do that. I just did a top and see that mono is using anywhere from 13% to 96% of processor. Wondering if that's normal? Ok, off to install ethtool.

[steve]

MX/mono uses 5% or less on my Pi 2, occasionally going a bit higher. It will use more when it's creating the files for uploading to the web, mine briefly touches 40%. It will use more on a model B, but I can't remember how much it used when I had it running on one of those.

[mzambo001]

I'm doing a clean install on the raspi 2 and reinstall of everything. Hopefully a fresh start will yield better results. If it starts dying again, I will go wireless and see if that works. If not, I'm going to a real sized computer. I love that the pi is so small and uses so little power but if I can't get it to work reliable, then what good is it.

[mzambo001]

Ok, got everything freshly installed. Did a new install of Raspian and updated. Installed mono the downloaded the latest beta of MX and got it running last night. Well the good news is that the network did not drop out overnight The bad news is that Cumulus crashed at 9:40am ish this morning. I could find nothing in the /CumulusMX/MXdiags logs or syslog. Not sure where else to look. I restarted it this morning. Let's see how it goes...

One good note is that the cpu usage is not as high as it was on the older raspi. It stays about 10~15% but will jump to 60% occasionally. I also turned up debug and turned on ftp logging although I don't think the ftp logging will tell me anything.

[steve]

If it actually crashed, there should be an exception report on the command line and in the diags file. But it depends what you mean by "crashed".

[mzambo001]

Nothing in the diag report. I have no command line because it starts via script in rc.local which redirects output to null. If it crashes again, I will log on with a monitor and keyboard and start it that way so I can see what happens.

When I said crash, I mean that mono is no longer running, nothing is updating and I can't see the gui at x.x.x.x:8998. Maybe it has something with the way I start it? Here is what I have. In rc.local a call a script.

Code: Select all

sudo /home/pi/startTheWX >> /dev/null 2>&1

in the script it is just this

Code: Select all

screen -amds bash mono /home/pi/CumulusMX/CumulusMX.exe

[steve]

Perhaps it's mono which is crashing, which might explain why there's no exception report in the diags log. It would perhaps be useful to be able to see the command line output.

[rogerthn]

Something like this might help
Remove screen from the script and in rc.local
/usr/bin/screen -fa -d -m -S cumulus sudo /home/pi/startTheWX

There might be other adjustments needed as well but the idea is to be able to see the command line output using screen -r cumulus

[mzambo001]

Good idea but I can't get it to work correctly. I decided to just take the batch out of it and try to start it up via command line. It runs but I can't get to the screen session. Here is what I entered.

Code: Select all

screen -fa -d -m -S cumulus1 sudo mono /home/pi/CumulusMX/CumulusMX.exe

And when I do a ps aux, I get

Code: Select all

pi        2393  1.6  0.4   6248  4472 pts/0    Ss   16:36   0:00 -bash
pi        2410  0.0  0.2   4560  1972 ?        Ss   16:36   0:00 SCREEN -fa -d -m -S cumulus1 sudo mono /home/pi/CumulusMX/CumulusMX.exe
root      2411  0.1  0.2   4812  2752 pts/1    Ss+  16:36   0:00 sudo mono /home/pi/CumulusMX/CumulusMX.exe
root      2415  7.4  2.2  40160 21260 pts/1    Sl+  16:36   0:01 mono /home/pi/CumulusMX/CumulusMX.exe

And I can't access it to see what the output is.

[rogerthn]

Try screen -ls as pi and as root to check where the "screen" is