DNS Broke without reason

Network DNS
1

Hello,

The DNS system keeps shutting down for no apparent reason.

The server check shows everything is OK.

dbs-verif-ok
dbs-verif-ok980×720 42.5 KB

I reconfigured the IP address of the RED interface (using the same IP), the “change” took effect, and the DNS is working again!!

Has anyone else had this problem?

Regards.

2

If you have enabled DNS Firewall, check if you have enough memory.

Look in the /var/log/messages file for “Unbound has died” after “kernel: Out of memory”.

See:

3

Hi Phil,
Thanks for the quick reply.
So the DNS firewall was enabled, and I only had 4 GB of RAM.
I’ve disabled it for now (by unchecking everything) and increased the RAM to 8 GB (the advantage of a virtual machine :slight_smile: )
I’m now keeping a close eye on whether it’s stable, and if it is, I’ll close this thread.
Best regards

4

Hi

Everything was working fine this morning, so I enabled the DNS firewall by activating the lists,

dns-selection
dns-selection966×620 47.8 KB

and the DNS server crashed a little while later.
So I increased the RAM to 10 GB.

I’ll see if it holds up.

5

In fact, you should avoid activating lists multiple times because each time you click “Save,” memory consumption increases.
You should activate all the lists you need only once (and restart); you shouldn’t need more than 8GB of memory.

6

That’s exactly what I did, but with 8 GB of RAM.

Apparently, with 10 GB of RAM, it seems to work.

As we say where I’m from, “Je touche du bois” (I’m knocking on wood) :slight_smile:

7

I am from the same country as you. :wink:

8

Hi
Update on this issue

It seems everything is working now, and 10 GB of RAM is the minimum requirement.

Merci de ton aide et à une prochaine fois peut-être. :saluting_face:

9

I tried your DNS Firewall config with my VM. Replacing ‘fast_reload’ with ‘reload_keep_cache’ ( similiar function without the memory leak ) I get a RSS memory usage for DNS of 1.54G. This means without the leak 3.5G of free memory would be sufficient (during fast_reload a copy of the state is needed). A system with RAM of 5GByte-8GByte should be able to run your very big lists.

The memory usage is equal after a restart, which builds the data base from scratch.
Means in the moment there are only two possible cases

  • doing fast_reload with minimal DNS outtime, but huge memory consumption
  • doing reload with some DNS outtime, but no memory overhead

The solution is to repair the memory leak bug of unbound, but I have little hope ( the issue is known for a long time, and I haven’t got an answer from them yet :frowning: )

10

Thank you for this information.
So, since my computer is used for work, I won’t make any changes (since it’s working fine), and we’ll wait for a fix, which still seems a long time coming.
Best regards

11

See also High memory consumption: DNS Proxy Server