On 2/17/24 00:54, Natxo Asenjo via FreeIPA-users wrote:
hi,

a bit late, but you should check the forwarding logs (maybe enable them, bit unsure if it is enabled per default on named).

Without any proof, my gut feeling is on dnssec :-), I have had to turn it off a few times.

Regards,

Natxo Asenjo

On Tue, Jan 30, 2024 at 5:11 PM David Harvey via FreeIPA-users <freeipa-users@lists.fedorahosted.org> wrote:
Just checking if there are any suggestions as to how to debug this effectively. The lack of smoking barrel log entries we've seen with it have left us a little stumped!
Thanks as always,
David

On Wed, 17 Jan 2024 at 10:54, Tania Hagan via FreeIPA-users <freeipa-users@lists.fedorahosted.org> wrote:
Hi Freeipa-users,

We are currently running Freeipa version 4.9.11 on Rocky 8.8. 

We have noticed over the last few months that external name resolution e.g. google.com fails to resolve on multiple Freeipa replicas even though the service named-pkcs11 remains up and running and journalctl or logs aren’t showing up any obvious errors to why this might be happening.  We temporarily fix this by restarting the service, but the problem comes back at random times. 

We currently have 39 DNS Zones

Our DNS Global Configuration has a forward policy of forward only, though the individual zones are set to forward first.

I’ve read a few articles that say maybe changing the forward policy might fix it, but nothing that mentions how to double check if changing the policy will fix it. 

Are there any useful troubleshooting checks I could run to either help explain why our service keeps failing at random intervals or confirm any changes would fix the issue without the risk of potential downtime of our DNS service?

Many Thanks,
Tania
--


If, after booting you wait an hour, and it seems to be normal for no good reason, maybe we're hitting the same bug.   I notice, each boot, the logs are filled with entries about 'notify' messages during which resolution requests time out.  My hunch is it's either some overlap of ip4/ip6 causing a race condition, or DNSSEC rr records holding some lock until they're all updated.  But that's just a hunch why turning dnssec off leads to normal resolution services.

Harry