Hi all, I’m running the latest eMMC flasher Angstrom image, kernel 3.8.13 on a BeagleBone Black. I’m using a 1A, 5V AC adapter. I also have an Edimax EW-7811UN N-wifi dongle, which is apparently rtl8192cu based. I installed the driver and added my network info to connman, and it successfully fetches an IP from my DHCP server. The dongle lights up when authenticated to my WPA network, and I can SSH into it and add/remove packages.
The problem usually comes when a large stream of text is being sent over SSH - ie compiling a large program, installing a package on NPM, or a installing lot of packages/dependencies via opkg. The SSH session will hang, and eventually time out with Write failed: broken pipe. The board will not respond to ping, or any other services running on it, it basically drops off the network. However, the light on the dongle remains on.
If, rather than running one of these commands in the foreground I start it with nohup, the process completes fine in the background with no loss of connection and I can verify the output with tails. Also, this happens on both Ubuntu 13.04 and Angstrom… Although I’m currently back to running Angstrom.
This hasn’t happened while plugged into the same router via ethernet, only when using the wireless. The board appears to be running fine, the heartbeat led continues and there is occasional activity on the cpu/flash leds. Won’t have an FTDI cable to confirm this until Tuesday, but if this is an obvious fix I’d like to take care of it sooner.
I will have a connection just fine then after several hours it will stop responding to all network traffic as if it were not there. The LEDs continue to flash as normal with cpu, heartbeat, etc… It’s quite annoying. I’m not sure if I want to even continue using it if I cannot create reliable projects.
Unfortunately I was never able to resolve this, although I didn’t try rebuilding the kernel or anything. I ended up switching to the Raspberry Pi, because while it’s a bit slower it seems to be much more stable for longer than a few hours.
I still have the same problem as well (on Debian); mucked with the net.ipv4.tcp_* parameters, still no joy in keeping either the WiFi or hard Ethernet ports alive more than a few hours. This is not just SSH - the ports become unresponsive to ping as well. Given my application requires availability, this is a problem.
The connections seemed to say alive longer when I set up a process to ping the router every 5 minutes or so, but still died eventually.
Does anyone have suggestions as to how we might monitor the ports to determine the cause of death? I’ve been netstat’ing, logging and ps’ing every way I could imagine, with no clue as to what is causing the problem.
Guess I could set up a cron job to kill the board every few hours, but that seems inelegant
I have had no luck with any small wireless dongle other than the ASUS N10 dongle. I've been able to keep relatively stable connections with it. Sometimes prolonged periods of intense activity like upgrading the system can seem to cause it to go off to la-la land never to be heard from again until after a reboot. That happens a lot less these days. Linux wifi has so many moving parts that it's hard to figure out which part blew up in your face. I'll say though I have tried various other wifi dongles that were recommended for BBB and all suffer exactly the fate you describe. at some point the network connection is cut and never returns until a complete reboot. So I just stick with these as they work out of the box with Arch Linux.
