Upgrade 2.4.0: firewall rule with alias and FQDN not working anymore

blueivy

What says Status/System Logs/System/DNS Resolver?

DNS Resolver only has the 'unbound' process. There is nothing of filterdns or dnsmasq in there. There is also nothing in System|General either for either filterdns or dnsmasq.

Are you not using DNS Forwarder service rather than DNS Resolver? I'm assuming there are different 'process' entries.

I'm happy to check anything else out to try and resolve this.

ggzengel

edit: This is forwarder, going to have to forward somewhere ;) I have not used the forwarder since they enabled unbound.. Well really before that when unbound was just a package. A resolver is just so much better than a forwarder. Not sure why anyone still uses it to be honest ;)

I tried to migrate to unbound last year but I had some problems: https://redmine.pfsense.org/issues/6065
Because I have more than 40 Overrides I don't like to try it again on this pfsense.
And there are still some unwanted effects with unbound: https://redmine.pfsense.org/issues/7884

blueivy

@johnpoz:

In a nutshell if you have an alias that is not working, you need to check the table. If entries not in the table then you need to figure out why the resolution of whatever FQDN is not working is not in the table. Pfsense needs to be able to resolve the FQDN you put in there for it to be able to put in the table..

So normally such problems just come down to name resolution troubleshooting.. Which doesn't look like any was done before bug report filed ;)

So what do you suggest beyond what has been done (by me)?

ggzengel

I have messages from filterdns in there. It's not dnsmasq and not unbound.

Even with unbound on an other pfsense I get this:


Oct 13 01:59:09 	filterdns 		adding entry 79.1.2.3 to ipfw table for host dummy.dyndns.org
Oct 13 00:59:06 	filterdns 		failed to resolve host dummy.dyndns.org will retry later again.
Oct 12 22:24:47 	filterdns 		adding entry 87.1.2.3 to ipfw table for dummy.dyndns.org
Oct 12 22:24:45 	unbound 	74165:0 	info: start of service (unbound 1.6.6).
Oct 12 22:24:45 	unbound 	74165:0 	notice: init module 0: iterator

ggzengel

I still had these two filterdns running.
I killed both. For the older one it was enough to send kill. The second one needed kill -9 to stop.

I removed the test entry with a FQDN inside and pfsense started a new filterdns.
Now it's not spamming any more. Only on changes (add/delete entries or changing IPs) I see filterdns entries in log.

blueivy

@ggzengel:

I still had these two filterdns running.
I killed both. For the older one it was enough to send kill. The second one needed kill -9 to stop.

I removed the test entry with a FQDN inside and pfsense started a new filterdns.
Now it's not spamming any more. Only on changes (add/delete entries or changing IPs) I see filterdns entries in log.

I added a new alias and a new FQDN (www.barrymanilow.com) and on query it's table (Diagnostics -> Tables) it has no entries. I get no filterdns entries in System|General or System|DNS Resolver logs.

filterdns does exist and is running (Diagnostics -> Command -> ps -A | grep filterdns).

I do get these errors in the System|General log file (whcih could have been therefore prior to the upgrade and they are maybe a red herring):

Oct 17 19:29:39 dhcpleases kqueue error: unkown
Oct 17 19:29:38 dhcpleases Could not deliver signal HUP to process because its pidfile (/var/run/unbound.pid) does not exist, No such process.
Oct 17 19:29:38 dhcpleases /etc/hosts changed size from original!

The /var/run/unbound.pid does exist.

I also did a 'cat /etc/hosts' and the nas.fqdn.private entry is in there. I think we can discount the 'if pfsense cannot resolve it it won't be in a table' issue as pfSense can not only resolve it, it's put it into it's Hosts file.

So the issue is that Aliases have no table entries.

And I'll say again what I said earlier the upgrade to 2.4 broke this.

blueivy

So I'll update the post with the fix for this.

The first DNS server listed in System -> General Setup was dead. It was working fine as there were anotehr 3 DNS servers in there. As soon as I replaced this one server with a working one I started seeing 'filterdns' entries in the System|DNS Resolver log. I checked the rules and they have started working now as well.

What I don't understand is:

1. The majority of the aliases are for internal IP's and therefore don't need external DNS resolution;
2. The majority of the aliases are for DHCP leases and are therefore registered by the DHCP service and appear in the pfSense Hosts file so again don't need external resolution;
3. If you have 4 listed DNS servers and one breaks then why should this stop aliases working;
4. What has changed that this issue did not appear before the upgrade to v2.4;

If I have a working system and I upgrade it and parts of it stops working then that's a problem. It's a bug. A bug in the upgrade. A bug in the way something works. But it's a bug. Something that should work doesn't. That's clear from this.

ggzengel

If I understand it right?

You have local DNS entries which appear in /etc/hosts?
And now the first of the external DNS servers is not responding and local IPs (from /etc/hosts) are not resolved in alias table?

Normally nsswitch.conf looks like:
hosts: files dns

What shows yours?

blueivy

@ggzengel:

If I understand it right?

You have local DNS entries which appear in /etc/hosts?

Yes. All of them were static DHCP leases. All of them resolved using Diagnostics -> Ping and Disagnostics -> DNS Lookup in pfSense.

And now the first of the external DNS servers is not responding and local IPs (from /etc/hosts) are not resolved in alias table?

They don't appear in the aliases tables. And filterdns entries were missing from the System|DNS Resolver logs.

Normally nsswitch.conf looks like:
hosts: files dns

What shows yours?

It is:

hosts: files dns

Exactly the same.

ggzengel

I think you should open a bug in redmine.

blueivy

@ggzengel:

I think you should open a bug in redmine.

I would do normally. But there is already a bug open for this. The way it was also completely dismissed without waiting for further information and pushed back to the forum means I'm not going to waste my time going through the hoops to do it. I appreciate that there was not a lot of information given on the issue raised but the way it was handled was poor. Pre-empting an issue as not a bug 'because we don't see it here' is a naive viewpoint and a does not encourage people to feedback on their project.

But I do appreciate your help in this ggzengel. Between the pair of us it lead me to find what I did. It's been much appreciated.

ggzengel

I restarted my pfsense and got only one filterdns and it's working.
Now I will have a look how long it will be stable.

ChrisCCC

Hi All,

I know this is an old topic, but I too, have noticed this issue occurring since an upgrade to 2.4.2. This definitely wasn't an issue previously, and very few config changes have been made since the upgrade.

I don't fully understand the process used to build these FQDN aliases, but I'll provide as much info as possible, in the hope it helps narrow down the root cause.

I've created a test Alias, called Host_Test, containing the FQDN 'www.test.com'.

Viewing the table entry for this alias shows an empty table.
DNS servers for the firewall are set to 8.8.8.8 and 8.8.4.4. DNS forwarders or resolver are not in use.
DNS resolution for this hostname is working fine for both DNS servers under status -> DNS Lookup.
Runninng 'ps -A | grep filterdns' shows there is a process running called filterdns.
If I view the log under System -> DNS Resolver, I can see that on the date of the upgrade (I assume on first boot after) there are entries such as the below, for all almost all FQDN aliases configured on the firewall. There have since been no events logged in this log.

filterdns failed to resolve host s186.fmp12-hosting.co.uk will retry later again.

This firewall has an HA partner, which doesn't seem to be experiencing the problem. Based on the total lack of logs since the primary firewall's initial boot, I'm wondering if the root cause is the process hanging (I assume 'filterdns' is the relevant process). Is it possible to safely kill and restart this process, or are there other considerations when doing this?

ChrisCCC

Quick followup. It looks like the process was hung. It's currently working after running "killall -9 filterdns" then saving and applying an Alias to restart the process.

What's potentially concerning is how soon after bootup this process seems to have stopped responding. Not sure if this is a one off for me, or something peculiar that's happening since the upgrade. I'll update this post if I notice the issue reoccur, especially after the next reboot.

Valeriy

I can confirm the issue and workaround by ChrisCCC
(Filter DNS service hangs, killall - 9 filterdns and then Filter reload (in pfSense GUI) solves the issue.)

I got the same problem after upgrade to 2.4

Currently running snapshot [2.4.3-DEVELOPMENT (amd64) built on Sun Jan 07 20:44:55 CST 2018]

Things to consider (in no particular order), that might be causing it:

I have substantial amount of hostname records in different Firewall aliases (hundreds)
After a while some hosts become obsolete (i.e. hostname does not resolve to IP address)
sometimes DNS servers might not be responding quickly (for such a big volume of DNS queries, perhaps)

I guess something is broken in filterdns algorithm after release 2.4: either after incorrect response from DNS server or absense of response causes it to hang.

snarfattack

Thanks for the workaround… this bug is driving me nuts too. Killing filterdns fixed the issue, at least temporarily for me. After a couple updates, it'll fail again.

Valeriy

It drove me crazy too, I wish I could have read this thread before I spent a few hours looking what is wrong.

Also,
https://forum.pfsense.org/index.php?topic=141441.15 is same topic.
Maybe Moderators can merge it?

echu2016

Does any body having this issue uses pfblocker ? In my case removing this PKG and replacing it by url tables aliases solved the issue.
Filterdns was receiving several sighup signals before, before hanging.

trees spanning the land

@Valeriy:

I have substantial amount of hostname records in different Firewall aliases (hundreds)

^This seems to be the problem, I had an alias with a list of Hundreds if not thousands of server Host names deleting them and then simply going to the other aliases with empty tables and clicking save to trigger a filter reload and update to the tables for those aliases has fixed the issue.

For everyone having this issue I'd recommend checking your Aliases to see if possibly you or someone else maybe added an alias into the firewall with a very large amount of hostnames and doing a backup of them and trying a delete and re-applying the other affected ones.

If we get some people coming back saying this is the cause then the devs will know what to investigate and hopefully get a fix put out for it.

I'm On 2.4.4-DEVELOPMENT If anyone finds that info Helpful.

qwertyface

I've just experienced this issue, and I think I have a solution.

For a while (including before the upgrade) I'd been seeing errors saying something like:```
There were error(s) loading the rules: /tmp/rules.debug:24: cannot define table bogonsv6: Cannot allocate memory - The line in question reads [24]: table <bogonsv6> persist file "/etc/bogonsv6"
@ 2018-05-16 19:13:38


From [https://forum.pfsense.org/index.php?topic=145990](https://forum.pfsense.org/index.php?topic=145990) I learned that the size of the BogonsV6 file grew substantially recently. The thread suggests increasing maximum table entries. While investigating, I noticed that the tables for my aliases were empty (**Diagnostics** > **Tables**). After increasing the maximum table entries, my aliases were showing up in the tables view.

I think this was probably a latent issue before the upgrade, caused by the size of the BogonsV6 table, and it's the reboot which has caused it, not the upgrade.