FTL stops running

Expected Behaviour:

Been running pihole on raspberry pi for a long time. Raspi 4 currently.

Updated yesterday.

Since have had numerous issues with DNS not running. Initially was using unbound but that would not run so it was uninstalled.

FTL kept quitting also so uninstalled pihole and reinstalled again today. Using only the local DNS options.

FTL keeps stopping.

I also have a second pihole built today on Unifi UDM Pro. This is also exhibiting the same behaviour with FTL continually failing.

Actual Behaviour:

DNS is not working.

Debug Token:

Raspberry pi: 35DVNTXf
UDM Pro: kLalDFTV

I just checked your debug token and found this line in FTL.log of your Raspberry Pi:

   [2022-08-05 18:00:09.839 5605M] FATAL: realloc_shm(): mremap(0x3f3d4000, 1923170304, 1923350528, MREMAP_MAYMOVE): Failed to reallocate "FTL-queries": Cannot allocate memory

Cannot allocate memory indicates that the memory on the machine has been used up. Your Pi-hole seems to be using almost 2 GB at this point. Looking at your query log, you seem to have created an infinite loop with your router asking your Pi-hole, your Pi-hole forwarding this to your router which then asks your Pi-hole ... This loop generates more than 5000 queries per second which your Pi-hole handles just fine until it runs out of memory.

When you resolve this infinite loop, the crash will be resolved.


Your UDM Pro shows a lot of errors in the debug log, like

*** [ DIAGNOSING ]: contents of /dev/shm
/dev/shm does not exist.

*** [ DIAGNOSING ]: contents of /etc
/etc does not exist.

which indicate that something is broken at the system-level (outside Pi-hole). There is also no FTL.log file on this device so maybe FTL does not crash here but, instead, just never starts at all?

We also see

*** [ DIAGNOSING ]: Operating system
[✗] Distribution unknown -- most likely you are on an unsupported platform and may run into issues. (https://docs.pi-hole.net/main/prerequisites/)

Thanks for the reply. Interesting about the memory as I saw disk getting filled up with log files after I posted too. Not sure where the loop is but I wonder if it's the two piholes not coexisting (as the UDM one is the router too, but on a separate vlan).

Not too worried about the UDM at this stage as it was more of an experimental to see if I could backup the pi which was causing issues.

Odd that the orignial unbound issue started after the FTL upgrade but maybe a red herring. No other issues reported with that version?

I'll kill pihole on the raspberry and start again from scratch I think. May also rebuild the UDM as that installation was fraught with errors due to out of date build instructions.

Thanks again.

I do think your current issue comes from a router misconfiguration leading to the loop we're seeing. I'd try that first before messing around with your installation in the Raspberry Pi. My best guess is that the upgrade forced a restart of FTL causing it to reread all relevant config files with might have been modified but this didn't actually change anything until the restart happened.

No even remotely similar issues have been reported to my knowledge.

Thanks. I wonder if the WAN DNS on the UDM pro has been set incorrectly somehow as that might send the request back to the LAN side DNS. I'll take a look now.

I found a related issue here where CF was causing a loop condition from the router forwarding requests to pi-hole DNS making millions of requests and causing FTL to crash due to RAM shortage.

I removed pi-hole addresses from my UDM Pro WAN DNS configuration and disabled Unbound (for now) and the issue resolved. I can still use CF in this configuration. UDM Pro passes on the DNS requests to my upstream DNS without issue anyway (according to DNS Leak Checker) so no need to configure at the WAN side it looks like (in fact a number of sites recommend against it due to unspecified issues - I know now what those issues are!).