USB ports stop working after a week, needs restart

[varesa]

I am working on a project with a friend that involves raspberry pies that run 24/7, periodically fetch data from a server and display it on the screen.

The issue we have is that apparently the USB ports stop working after about a week of uptime. The same issue happens with five different units.

The only device plugged into the RPis normally is a 3G dongle through a powered USB hub. I notice the issue from server logs that a certain unit has stopped updating the data. When I drive to see the device, it is still running and looping the data on the screen but I can't connect a keyboard. I assume that the USB controller or something related to that stops working.
If I restart the unit everything is fine.

The raspberry pi is revision 000e (v1 model B, 512MB, PCB rev. 2).

I am in the progress of updating the image from Wheezy to Jessie. It is possible that this was fixed in some kernel update, etc. If I had two extra units I'd just rpi-update one and leave it for a week while working on the upgrade on the other.

Does this sound like something that has been fixed or is possible to fix in software or is there some other likely cause?

[RaTTuS]

no idea , newer will help IMO
or you could just issue a system reboot if it looses 3g connectivity or just reboot it every day ....

[hansotten]

Possible causes might be an application/driver/deamon with a memory leak, slowly eating all available and choke the system.
Can be is the OS, or in the applicatiosn you have added.
Can you monitor remote/log the system resources?

The easy way out is a reboot every day via a cron job.

[varesa]

Yes, the reboot is my planned workaround. Since the devices are in a publicly visible spot I'd prefer to reboot them during the night instead of showing linux boot messages too often I just have to force a certain HDMI output mode since the display gets powered off for the night with a clock switch and the pi can't detect resolution/refresh rate.

Running out of RAM is something I haven't even thought of but could very well be possible. I will check it out on the unit I have.

Also one of them has crashed again. This time I asked the friend to swap out the card and bring me the crashed one. I might be able to see something in the logs (dmesg, syslog, kern.log, etc.)

[varesa]

I finally got the crashed card.

dmesg: http://paste.fedoraproject.org/345235/58930922/

kern.log: http://paste.fedoraproject.org/345236/45893092/