Lose internet every 14 days

[bigjohns97]

Does anyone else lose internet every 14 days with this deployment?

I have setup a cron job to restart the service every week to work around this issue but I am just wondering if anyone else has this 14 day issue with the standard install guide?

[evie-lau]

I personally haven't run into this, with a current uptime of ~60 days. I'll leave this open in case anyone else chimes in with similar issues

[bigjohns97]

Thanks I appreciate it.

Full story.

I recently switched from a home built PFSense box where I was doing bypass to a Unifi UDMPM.

At first I just did the certs and WPA service and noticed everything was working so I didn't do any of the VLAN and MAC address stuff.

Then I had the first disruption right at 14 days exactly so I thought well let's do the VLAN 0 and MAC stuff it must be that.

Then the exact same thing happened right at 14 days again so I am kind of lost at this point.

What I think I will do is exactly what you have in your instructions (when I did the MAC clone I actually did it in the UI and saw it represented in ifconfig so I never did the script option in your instructions)

Maybe this will be the difference.

Thanks again for the write up!

[evie-lau]

Did you ever figure out the underlying issue?

I'm thinking it's unrelated to the setup in this guide, but I want to make sure it's out of scope before closing this issue

[bigjohns97]

I have not figured this out but ended up doing a workaround by adding

ctrl_interface=/run/wpa_supplicant

to the wpa conf file and then creating a script to run wpa_cli logoff and logon once a week to refresh the WPA session.

so far this has been my best solution to keeping my service from being disrupted every 14 days.

[bigjohns97]

Another odd thing I noticed is whenever I set the MAC via the script and not in the UI and reboot the MAC isn't implemented.

I am guessing this is because the system boots up, the interface comes online and the script runs and then gets overwritten by the unifi software refreshing and bringing down it's config.

I think this is probably different for you @evie-lau because you are using a UXG which is an independent gateway while I am using a UDMPM which is a controller/gateway/switch all in one.

[gpz1100]

I've been experiencing this on pfsense for nearly a year now, ever since I switched to using the odi dfp-34x-2x2 sfp module. At one point I went back to using the G-010G-A ont but the issue continued.

After much troubleshooting I discovered that connectivity would restore on its own following the next dhcp renewal. That is, dhcp renews every ~30min (1hr lease). This outage would occur sometime in that 30 min window. No connectivity remained until the firewall did the next automatic dhcp renewal. Say outage occured 10 minutes after the last renewal cycle. No internet for another 20 minutes.

Anyways, I set up script that gets activated when gateway monitoring detects no response. Script verifies a few things then forces a manual renewal. Outages now last < 30s.

@bigjohns97 I will test out doing the weekly logoff/logon. It'll be great if that fixes it permanently. My next scheduled outage is literally a week from today.

[bigjohns97]

I finally have xgspon in my area so I no longer have to mess with 802.1x. Sorry I won't be able to participate in the festivities moving forward.

[gpz1100]

I did the wpa_cli logoff/logon a week ago as well as several days ago.

Reviewing pfsense logs from this morning, sure enough, it lost connectivity literally 14days+20s since the last outage. So refreshing wpa didn't solve it on this end.

@bigjohns97 Specifically what service are you restarting? What's your cronjob and/or script look like? Thanks!

[bigjohns97]

Sorry @gpz1100 I moved over to XGSPON which doesn't require 802.1x but looking back on my comment

It looks like I added the ctrl interface variable to the wpa conf file and then used the wpa_cli to logoff and logon the wpa process refreshing the token which also should in turn refresh the dhcp lease

https://linux.die.net/man/8/wpa_cli

I'm sorry I don't have the scripts anymore because what I stated above but you should be able to use the ctrl interface as well as the wpa_cli commands to create a refresh script that either runs daily or once every 14 days. (this is what I was doing)

[gpz1100]

I've always had the ctrl interface, that's what allows wpa_cli to communicate.

It sounds like on your end doing the logoff/logon also forced a dhcp refresh? I'm pretty sure i've tried manual refreshes before but it had no effect on the 2 week timer.

If I recall, I have the gateway monitoring configured so that it needs to be down for 20-30s before an 'outage' is registered, which would then kill the dhcp client and cause it to refresh anew.

At this point it's not that big of a deal. If the skew each week is only 20-30s, then it will be a long time before it shifts to a weekday time when it becomes more than a nuisance.