Magic something issue (and more) on 2.7.0 - feedback/report

[senseivita]

Hi all, (though if any devs around, better)

I installed 2.7.0 a couple of days ago, twice. I was going to post on Redmine but my email is still offline so to avoid the risk of needing confirmation, I posting here.

1

First thing I did was changing the keyboard layout because it won't stick. But as soon as I did I got this error:

…and the layout didn't change, or not until the next session but not for the selected but one close enough, it seems. Then I remembered this tip of setting it using the keymap variable in /etc/rc.conf.local, it's still a little off but at least I don't have to face the error.

2

Also, when the systems boots up, there's some character/video-related glitches:

(If you know how to fix this—a package install like the localectl module on Linux maybe—please, tell.)

3

️ +1 for PPPoE issues.

Specifically, username/password appeared (in the error message) wrong, but they randomly not only failed on pfSense but also on Mikrotik CHR, VyOS, OPNsense and the ONT itself (in router mode).

After researching on the wrong thing, non-ISP-issued ONTs on Amazon, I stumbled on this this dying gasp feature on the description of a device; Modems and ONTs with this feature send a disconnect signal when there's sudden power loss so the credentials/session session is free to reconnect.

Since there were session-related errors which are always ambiguous/vague, I speculated myself an answer: maybe there a little tiny oversight in the code that kills the session/daemon/whathaveyou ungraciously. It's a stretch, but either I settled with that or called tech support and I'm too young to die due to an anger-induced cardiac arrest.

4

Next on the list is package re-installation. (Subtitle Does it really need to happen at once?)

I use OSPF, I'm new to it so I lack the experience to know if you're supposed to have static backup routes or not. I thought not since that's soft of the thing it's supposed to replace, but what do I know.

Routes/gateways are discovered and set up automatically and they're ready by the time pfSense finishes booting up, tunnels and all, but in addition to when you restore a full backup, I found out that on updates too pfSense wipes all packages before even getting the new ones to replace them. Despite being online, it still couldn't connect bc the DNS servers are two routers down and themselves depend on pfSense announcing itself to connect out. It's a catch 22.

(In an AD environment,) since DoH was maintreamed, DNS is at least half of my network infrastructure. Modifying the ruleset to allow pfSense get its own DNS if briefly could would get ugly, but I thought on a much quicker and less destructive fix: just enough high-metric static routes to reach the nearest DNS server. OSPF's metric is 110, I don't remember about RIP, but 150-ish should be fine so it can be left (or forgotten) in place without overruling dynamic routes when they came back. For all I know doing this setup is the norm and I just stumbled into it.

5

The top menu/bar/header when there are warnings present (see next) and it's set to show the hostname, it becomes 3 rows. I think that's unintended.

6

Last issue was this warning/crash: PHP_errors.log.txt

I thought I had disk corruption, this is when I reinstalled straight to 2.7.0 as opposed to the update before it.

The error appeared when it finally begun installing packages, though it seems it fixed itself later. I'm not sure when because I was distracted by a few other issues on OpenVPN, still am. I restarted a few times since. Fixed or not I don't want the firewall gods to smite me unblocking Windows Update on my servers or some other destruction from within scenario so I'm reporting it nevertheless.

Hopefully this is useful to somebody.

[stephenw10]

Hmm, that php error looks like this: https://redmine.pfsense.org/issues/13562

But that should have been fixed. You might want to add your result there.