@DL6ER Not really sure about that. But its not like I am sitting in front of my pi checking every DNS query. I am in the mid of my diploma exams, but when I find the time I will try to dump pihole -t output to a text file to see if there is any relation to the DNS queries
For how long have you been running DoH now? Somtimes it took many hours before it crashed.
This is the setup where I always have crashes:
pihole 5
DNSSCEC enabled
DoH enabled
Cloudflare as DNS provider
As soon as I disabled DoH everything works fine again.
I had this issue repeatedly with Cloudflared’s DNS over HTTPS. It’s what prompted the switch to unbound. As @yubiuser recommended, temporarily change upstream servers and see what happens.
For sure, and I hope nobody is wasting their time for such a thing, however, it is for what log files have been invented for
Check /var/log/pihole.log on the next crash (if you have logging enabled).
Almost 24 hours by now. So far, everything is still running, no crash. That's why I was asking for a specific query, maybe containing a reply violating the DNS standards and, by this, causing trouble somewhere where the error checking is to be improved.
Thanks, it's still running for me, hence, I'm fairly sure that
and my "normal" web browsing using nothing else than Firefox and Thunderbird accessing the web isn't sufficient to trigger the crash. In the past, we have often seen similar crashes in networks running Windows machines making a lot of queries to some Office domains (for telemetry stuff, I think). Do you use Windows/Office in your network? This, for instance, is something I cannot easily reproduce.
Yes due to my job use a windows machine. Especially since I am forced to use MS Teams I noticed that the microsoft domains dramatically increased. Especially mobile.pipe.aria.microsoft.com is going crazy during MS Teams usage:
By the way, DNS crashed again after I went to bed yesterday. The thing is I don't know when it exactly crashed, so I dunno which DNS queries could possibly cause it. But for now I will just stick with Cloudlfare without DoH as this seems to work. And as long as no one can reproduce the problem, I guess it is specific to my environment.
EDIT: A little sidenote. If someone tries to reproduce this problem I also have the following settings enabled in the DNS tab:
Maybe this is also pi related. As the webinterface is always super slow when this happens. Also I cannot even reboot with sudo reboot. Pi goes black but never comes back up. I have to actually unplug it from the power supply.
Hmm, yes, thanks for providing all the details here. So all crash reports seem to happen only with Microsoft products of various kinds, so far. And a very large fraction (maybe all, the information is not always there) is also related to using Cloudflare, maybe with, maybe without DoH.
The entire behavior of your Pi-hole is very strange (the need to unplug from power the hard way, etc.), I cannot really say how I feel about this. I don't see us being able to offer any real advise at this point.
I again looked at the crash log you posted on May 26. However, I'm not even sure if you encountered a real crash afterwards. It more looks like everything got viscous like hell without real crashing, especially because
This all leads me to two hypotheses at this point:
The DNS Online/Offline check may give wrong output if the DNS resolution is just slow (too strict timeout?)
Such laggy behavior on Raspberry Pis are typically related to memory issues. So, maybe, either your RAM or your disk (maybe a subpartition like /var) runs full. Maybe free -h and df -h can show something where the usage is extraordinary high.
This also brings me to the idea that it may not be a Pi-hole issue at all (?). If DoH is responding slow and eating up all resources on the Pi-hole this could very well lead the the very same symptoms. And this hypothesis actually fits better to the observation that the issue disappears when you take DoH out of the equation.
I agree with your conclusion. Funny thing is that it used to work without any issues (also using MS products in combination with DoH) in the past with V4. I will simply wait and hope someone stumbles over a solution somewhen Until then I will simply disable DoH. But still: Many thx to everyone, and especially DL6ER, who took the time and effort to look into this problem. Really appreciated the conversations and support!
Not sure what you mean with DNScrypt. If you mean DoH: From my personal experience I know that HTTPS traffic (especially with DoH) needs more resources and hence can use up more memory and storage for key exchange etc. I cannot proove it, but my gut feeling tells me that something in the chain of DoH floods my memory and therefore somehwen causes the issues mentioned above. Another thought of mine is if you have alot of queries to the same domain, like in my case M$ domains, that Cloudflare simply denies resolving as it is interpreted as "bullshit traffic". I know this must sound very high level from a developers perspective but I am just a IT system engineer doing his best trying to understand the problem and help solve it.
