if_pppoe problems with php-fpm causing loops. (resolved)

chrcoluk

@w0w I will do both stacks, it was something I did very quickly and I want to check if its still 12 bytes for ipv6 as well.

Patched it now. So on my 1500 bytes MTU, 1448 MSS for IPV4, 1428 MSS for IPv6.

w0w

@chrcoluk
I did the same and even captured WAN packets to confirm. Looks like it's working — we'll see. By the way, I'm using gigabit PPPoE.

stephenw10

Hmm, interesting. I can't say I've noticed that. But also I wasn't looking for it specifically.

chrcoluk

After some more testing, it looks like if_pppoe resolves an issue related to fragments that existed in the legacy pppoe code.

w0w

@chrcoluk said in if_pppoe problems with php-fpm causing loops. (resolved):

issue related to fragments

What issue?

chrcoluk

@w0w I had a device that had issues with small tcp packets, it still fails on the legacy code but now passes on the new code. I didnt really consider it an issue pppoe side before, but the issue is gone on if_pppoe.

chrcoluk

PPP session was down for a while late afternoon, seen notification from ISP, and then when I got to firewall could just see repeated "received unexpected pad0" on console, and logs looked like was in a loop retrying.

After I confirmed ONT ok, I disabled WAN, enabled it, selected apply and it woke up within seconds.

Now there was external cause for the initial disconnection, was an outage on CityFibre the FTTP local provider, so the issue is it failed to connect automatically rather than the initial disconnection.

stephenw10

Hmm, interesting. Did you try just disconnecting and reconnecting in Status > Interfaces?

ajtuk

I had the same issue today after CityFibre went down. The PPPOE connection does not restart, luckily I can SSH in from the FTTC line and reboot it. Then it works fine. But if the ISP drops the connection, it's either access the GUI and click Connect or reboot.

chrcoluk

There was a second later outage, on the second outage it came back up by itself.

chrcoluk

@stephenw10 I didnt try that as I had experiences recently, using that method could leave services down, whilst toggling in interfaces always seems to keep everything running cleanly.

There is a chance of more drops as AAISP havent closed the incident, given CF are refusing to communicate with them on what has gone wrong, if it gets stuck again, I will try a cycle on the interfaces screen.

chrcoluk

@stephenw10 It went down again, didnt auto recover, status interfaces said it was still up (no ip's), I took it down then back up again and it came back right away.

w0w

@chrcoluk
When the PPPoE interface shows 'up', it means it's in the process of connecting. How long did you wait? Next time look what does

pppcfg pppoe0

show.

chrcoluk

@w0w It was already down for 2 hours and I had people moaning at me, no time to sit there watching it I am afraid. I wasnt asked to sit and watch it either, just to restart it on the interfaces page.

But I will try and remember that command. Thank you.

Future testing at more convenient times might be possible via yanking a cable.

What is logged during what seems to be a connection attempt before I intervened, I removed all noise like service restarts etc.

If I remember right things like timeout settings used to be tunable?.

Jul 26 06:11:28 	kernel 	222645 	if_pppoe: pppoe2: LCP keepalive timeout
Jul 26 06:10:40 	kernel 	222596 	if_pppoe: pppoe: GENERIC ERROR (errortag 1)
Jul 26 06:09:46 	kernel 	222542 	pppoe2: link state changed to UP
Jul 26 06:09:46 	kernel 	222542 	pppoe2: link state changed to DOWN
Jul 26 06:09:46 	kernel 	222542 	pppoe2: link state changed to UP
Jul 26 06:09:38 	kernel 	222535 	pppoe2: received unexpected PADO
Jul 26 06:09:38 	kernel 	222535 	pppoe2: received unexpected PADO
Jul 26 06:09:38 	kernel 	222535 	pppoe2: link state changed to DOWN
Jul 26 06:09:38 	kernel 	222535 	if_pppoe: pppoe2: LCP keepalive timeout

chrcoluk

@w0w I am seeing indications in the log it isnt going down as you suggested, as an example dpinger stayed running on static pid, logging almost every second ping failures. Also no switch of gateway, when the switch criteria is "down" state. (my phone was left in all night as a internet uplink).
If its failing to go to a down state, it would make some sense as to why a manual restart works and leaving it alone doesnt. As the obvious question is what is different between me restarting it manually and it trying to restart itself, there must be a difference.

When I intervened dpinger exited on signal 15.

w0w

@chrcoluk said in if_pppoe problems with php-fpm causing loops. (resolved):

But I will try and remember that command.

~~pppcfg pppoe0 -d~~

~~Will be better as it return debug details .~~

Sorry, @stephenw10 gave the correct answer about the debug option — I somehow mixed it up in my head with something else.

@chrcoluk said in if_pppoe problems with php-fpm causing loops. (resolved):

If I remember right things like timeout settings used to be tunable?

When the last time I searched I didn't find anything related parameters, but maybe it's changed already, IDK.

@chrcoluk said in if_pppoe problems with php-fpm causing loops. (resolved):

Jul 26 06:09:46 kernel 222542 pppoe2: link state changed to UP
Jul 26 06:09:46 kernel 222542 pppoe2: link state changed to DOWN
Jul 26 06:09:46 kernel 222542 pppoe2: link state changed to UP
Jul 26 06:09:38 kernel 222535 pppoe2

I don't know is it expected behaviour to switch link state that fast, but it could be a reason for dpinger not going down and all other things happened.

stephenw10

Hmm, if it happens again try checking the status shown in: pppcfg pppoe0

You could also try enabling debug: ifconfig pppoe0 debug. But that will spam the console/logs at lot! Be ready with ifconfig pppoe0 -debug

chrcoluk

A retry delay, might be an idea? With the timing issue on ipv6 as well, I wonder if things just move too fast.
Because I dont understand how the new PPP works, from what I can tell it doesnt even have a config file, the parameters are just supplied directly, its hard for me to try and patch something myself to find a solution, I am hoping netgate are looking into it as its just me and ajtuk from one ISP.
Just in case it might be relevant, the WAN connection does go through a WAN side VLAN, a requirement from CityFibre.

This is also confirmed on the command output here.

# pppcfg pppoe2
        dev: igc1.911 state: session
        sid: 0x44a8 PADI retries: 0 PADR retries: 0 time: 02:13:26
        sppp: phase network authproto auto authname "x@x" peerproto auto 
        dns: 217.x 217.x

I will of course do as you said stephen, the incident is now closed, but if needed I will try to emulate it with a cable pull test, but I have someone relying on the uplink almost around the clock, so might be a little bit of time to get a chance to do it.

I am also prepared to have a conversation with AAISP to see if they can assist from their side, maybe they can notice something wrong in their logs.

stephenw10

Mmm, the VLAN should not have any effect. Many ISPs require that..

Can you replicate this manually by just reconnecting the WAN? That at least would be something we could try to replicate locally.

chrcoluk

@stephenw10 I thought about how I would do the test, I think best way is cutting power to the ONT, s the box will detect the carrier link going down if I pull the ethernet, I will try to do this test as soon as I can, but to try and mimic the conditions, I will probably need to keep at down for at least 15 minutes.

I will do this in about 5 hours or so, I think either killing the ONT power or removing the cable, will make the cable register as disconnected, I am not prepared to remove the fibre cable, as it leaves it exposed to dirt.

AAISP offer a kill switch for the PPP, and the last outage was caused by them doing a "admin reset" it was logged as that on their logs. I guess to kick me of the LNS for rebalancing purposes, so I will get online via my phone and use the PPP kill switch for the test, this leaves the ONT powered up and ethernet cable in a connected state to mimic the failure conditions.