Insanely weird connection issues with LTE modems (but issue is with TO)

I would download the /etc directory from the router to another device prior flashing. Whilst it should not be restored recursivly it may make it easier to get back up to speed after the flashing.

To be honest I’d rather start from scratch. Just to be sure there’s no leftovers at all.
OK, medkit in progress. I’ll update status here soon.

Update flashed latest medkit. Plugged my USB modem/router, system automagically registered new usb0 ethernet interface; no packages required. I’ve added it to WAN fw rules. Got connected to internets. Issue is happening again:

This is my simple test bash loop thing:

while true; do
  timeout -k1 1 dig @1.1 &> /dev/null  && sleep 2 || \
    echo FAILED \@ `LC_ALL=en_US.UTF-8 date`

And this is the output of it whenever the issue happens:

FAILED @ Fri May 25 17:12:26 CEST 2018
FAILED @ Fri May 25 17:12:27 CEST 2018
FAILED @ Fri May 25 17:12:28 CEST 2018
FAILED @ Fri May 25 17:12:29 CEST 2018
FAILED @ Fri May 25 17:12:30 CEST 2018
FAILED @ Fri May 25 17:12:31 CEST 2018
FAILED @ Fri May 25 17:12:32 CEST 2018
FAILED @ Fri May 25 17:12:33 CEST 2018
FAILED @ Fri May 25 17:12:34 CEST 2018
FAILED @ Fri May 25 17:12:35 CEST 2018
FAILED @ Fri May 25 17:13:10 CEST 2018
FAILED @ Fri May 25 17:13:11 CEST 2018
FAILED @ Fri May 25 17:13:12 CEST 2018
FAILED @ Fri May 25 17:13:13 CEST 2018
FAILED @ Fri May 25 17:13:14 CEST 2018
FAILED @ Fri May 25 17:13:15 CEST 2018
FAILED @ Fri May 25 17:13:16 CEST 2018
FAILED @ Fri May 25 17:13:17 CEST 2018
FAILED @ Fri May 25 17:13:18 CEST 2018

Note: it corresponds to actual outage and is just a visual representation to me. All the sympthoms I’ve described earlier are also happening at that time (that is lack of ability to establish new TCP session etc)

I’m going to update this thread with my further findings. My next plan is to find (if possible) medkit from Q1 2017 and try it instead of latest 3.10.

The Turris team seems to be more interested in selling “new” products like MOX rather than solving problems of the existing products like Omnia. After more than a year playing around (as a regular user) with the Lte modem on Omnia, I concluded that Omnia is not reliable product for use with LTE modems. Among other problems some packages (like mwan3) are obsolete in the turris os and no one care to update packages (The turris team also seems to never engage in forum issues related to the Lte modem).

With almost each update of turris os, I am facing connection problems with the Lte modem. After TO 3.9, resetting the modem with uqmi is no longer an option and one needs to unplug the omnia to reset the modem (in my case not an option).

With the Lte modem provided with Omnia, I also had Lte problems which were related to conntrack, even if connection was established I could use the connection only after conntrack was initiated, which could happen long after the connection to the provider.
An other source of problems may be mwan3 if used. Depending on the mwan3 settings you might not be able to use the Lte connection before mwan3 decides that the interface is up. I have not tested TO 3.10 yet as I am not able to remotely reset the modem.

Try on IRC. @Pepe is super helpful and in fact everything is now working fine to me. Issue was NEVER related to LTE modem per se, and honestly I was mistaken when I was setting the title of this thread.

To everyone. Original issue seems to be gone (that is regular hurly thing). However I am still facing random occurrences of this problem like 2-3 times a day, not regular, and in fact it is rare enough for me to give up on further troubleshooting. If I didn’t have the short script I mentioned earlier running 24x7, I wouldn’t even know there’s any problem.

Solution apparently was to flash TO with 3.10. I’ll update the thread if I find anything new to share with you guys.

Also extra note. Current 3.10 has ALL the packages and kernel modules preinstalled for both of my modems, so I didn’t have to pull a single extra package. That’s sweet!