Lost Connection to API

FL-TOAST · October 19, 2019, 11:04am

Please follow the below template, it will help us to help you!

Expected Behaviour:

P:iHole should "work" reliably -- as it has for two years

Actual Behaviour:

"Lost Connection to API"

Debug Token:

Unable. I ran the script via the web interface and it appears no inforamtion was uploaded:
[✗] There was an error uploading your debug log.

- Please try again or contact the Pi-hole team for assistance.*
- A local copy of the debug log can be found at: /var/log/pihole_debug.log*

I tried to deal with this several times via updating the OS, running "pihole -up" an rebooting. No effect. PiHole runs for a few hours and then the message shown above is seen. Running on a RPi, a 3B, I think. No power issues -- running on a ferroresonant UPS.\

Any interest/use in diagnosing this? Should I just start over?

jfb · October 19, 2019, 11:15am

This will temporarily reset the nameserver on the Pi to bypass Pi-Hole DNS.

sudo nano /etc/resolv.conf

edit nameserver 127.0.0.1 to nameserver 9.9.9.9 or your preferred third party DNS service, save and exit

Run pihole -d and upload the debug log

FL-TOAST · October 20, 2019, 11:10am

Thank you very much. Token is: asbx8dr4ur.

FL-TOAST · October 22, 2019, 10:42am

Token is in message sent a couple of days ago, above. Although the problem occurred several times nothing further observed over the last two days. Just disregard it? - Rick

FL-TOAST · October 23, 2019, 12:40pm

Happened again. "Lost connection to API." From what I can see, the uploaded file is automatically deleted after two days so the token I identified above is not of no value. Right?

Since there was no response to the previous upload, should I do it again, or just start over?

Edit: New token is pzcnmvikul. Could someone take a look at this?

FL-TOAST · October 24, 2019, 12:14am

OK, could SOMEONE give me some help here, please? Someone?

Are the debug logs I uploaded of any value?

jfb · October 24, 2019, 1:34am

Apologies for the delay. We're all volunteers and I was out of town the last few days.

I'll have one of the developers look at the error in your log from the FTL crash.

FL-TOAST · October 24, 2019, 10:16am

No Apologies needed. I'm always amazed at what you guyz do.
PiHole "broke" three times within the last 24 hours. I have two others that have been perfect and this one was also for many months. So, I guess this is an outlier.

TU!

Rick

DL6ER · October 24, 2019, 2:17pm

This seems to be an error in the embedded dnsmasq code. Did you enable DNSSEC validation in Pi-hole?

Please provide the output of

addr2line -e /usr/bin/pihole-FTL 0x40eea 0x33c72 0x4b22e 0x4aff8 0x13c5a 0x1a25c

FL-TOAST · October 24, 2019, 11:50pm

Hi. Since I can't copy between Firefox (or the other way) and RDT I suppose I could have made an error in doing this "manually," but it appears the result is:

/root/project/dnsmasq/forward.c:1050
/root/project/dnsmasq/dnssec.c:887
/root/project/dnsmasq/dnssec.c:1923
/root/project/dnsmasq/dnssec.c:304
/root/project/signals.c:46

Does the following image answer your question re DNSSEC?

I've not touched the settings of this PiHole for a couple of months. Only within the last few days has the "lost connection to API" issue occurred.

jfb · October 25, 2019, 1:21am

No. There is a selection box on the lower part of the same screen titled “use DNSSEC.” Is that checked or not checked?

FL-TOAST · October 25, 2019, 10:40am

Good morning. Yes, that box is checked.

jfb · October 25, 2019, 12:11pm

Uncheck the box.

FL-TOAST · October 25, 2019, 12:17pm

Done. TU.

OK, now a question: I have three PiHoles, each at a different location. All are configured the same -- using Quad9's DNSSEC resolvers and that box has been checked. Two of the three have operated flawlessly. The present one also was working perfectly until a week or so ago.

The "DNSSEC resolver test" on the same page shows the desired result in each case.

I'm confused as to "what happened." I'm wondering how the others have survived if they are mis-configured and if the "Use DNSSEC" box has any purpose if DNSSEC is disallowed.

Maybe you can clear up the "fog?" ;<)

DL6ER · October 25, 2019, 1:57pm

Looking at the code lines you provided, it seems like dnsmasq crashes when it reaches the code that buggle sorts the RRset into the canonical order. Could you just one more time check if the line numbers you typed after the : are indeed correct?

DNSSEC is meant to be used, that's not the point here and @jfb just asked you to quickly uncheck it to prevent any further crashes for you.

Could you share also the content of your /var/log/pihole.log at a time the crash happened? It may be that this bug is still existing and it would be good to know which domain was causing the issues to be able to reproduce it locally (in order to be able to fix it).

FL-TOAST · October 25, 2019, 3:51pm

Hi Nick. Yes, I double-checked. Seems OK.

OK. That explains @jfb's request. Logical -- and thanks!

Not sure how to share that file with you. It appears to be quite large. Not sure a copy/paste will be the best approach. I'm not a LINUX guy. Your advice as to how I may best accomplish that would be greatly appreciated -- and I'll try to do so promptly.

I appreciate your assistance with this!

DL6ER · October 26, 2019, 4:42pm

I'm not sure how to do this best. Maybe you can switch DNSSEC back on and let
http://pi.hole/admin/taillog.php
run in a tab on your browser. I have never tried if this works, but it might.

When you loose the connection to the API, the most recent (few) lines are of interest.

FL-TOAST · October 26, 2019, 5:26pm

OK. Let me try that!

system · November 16, 2019, 5:39pm

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.