NTP Not Working [SOLVED (totally)]

[beremonavabi]

@johnpoz said in NTP Not Working [SOLVED]:

You sure do not need to be listening on wan for it to go outbound and talk to the ntp servers.

Well, that's what I thought when I set it up initially (and it worked that way at the time). But, all I can say is that, at this time (with my current configuration on pfSense 2.4.3-Release), without WAN being selected, the NTP service doesn't work. Merely adding the WAN interface under Services > NTP causes it to immediately start working again.

EDIT: Another odd thing is that according to that quote I posted from the other thread, above, the other solution is to have the selected interface with the lowest IP address have outbound NAT configured. Well, for me, that's my LAN interface (192.168.1.0.24). Firewall > NAT > Outbound for that interface looks like:

Interface	Source	Source Port	Destination	Destination Port	NAT Address	NAT Port	Static Port	Description	Actions
		WAN	192.168.1.0/24	*	*	*	WAN address	*		LAN to WAN

So, unless I'm mis-reading that, I do have outbound NAT configured for it. Odd.

[kpa]

It's not as simple as you would think. NTP service uses UDP only and there is only one listening socket for sending out queries and listening for replies on UDP port 123. If you set the service not to listen on your WAN interface that will also cut off the service's ability to send anything out on the WAN interface. Yes I know, DNS (for example) does the right thing and uses a separate socket for sending out the queries with a randomized source port, unfortunately NTP hasn't caught up with the new developments yet.

[jahonix]

So you're saying all my setups with NTP servicing all internal networks but WAN don't work? I must have missed something then...

[kpa]

Look at your outbound NAT rules. There should be a rule for outbound NAT'ing anything sourced from 127.0.0.0/8. That is the thing that makes your set up still work if you happen to set the service not to listen on the WAN, the service will instead use the localhost 127.0.0.1 address for the traffic and it will be properly NAT'ed.

Edit: The service might also select your LAN interface in case WAN is not available and if the LAN interface has a routable IP address it works just fine, also if the LAN IP address is an RFC1918 address you're still good because of the standard outbound NAT.

[beremonavabi]

@kpa said in NTP Not Working [SOLVED]:

Look at your outbound NAT rules. There should be a rule for outbound NAT'ing anything sourced from 127.0.0.0/8. That is the thing that makes your set up still work if you happen to set the service not to listen on the WAN, the service will instead use the localhost 127.0.0.1 address for the traffic and it will be properly NAT'ed.

Interface	Source	Source Port	Destination	Destination Port	NAT Address	NAT Port	Static Port	Description	Actions
		WAN	127.0.0.0/8	*	*	*	WAN address	*		Localhost to WAN

I've got that rule, too. But, without WAN being selected for NTP, it still didn't work.

[johnpoz]

I do not have wan selected, and ntp works just fine to outside ntp servers.

And yeah your loopback should be outbound nat.. Unless the user turned off automatic outbound nat then yeah loopback would work and or any of other lan side interfaces. If they turned off loopback nat, pretty sure that would break a bunch of stuff.

[cyberzeus]

This post is deleted!

[beremonavabi]

Just to double-check this (in case my forcing a clock update in my troubleshooting attempts "fixed" this), I unselected WAN from the selection list on Services > NTP and restarted the NTP service. I was right back in the same situation I was when I started this thread:

The NTP Status Dashboard widget has the time, but lists the Sync Source as “No active peers available.”

Under Status > NTP, all the pools show a status of “Pool Placeholder” and non-pools show “Unreach/Pending.” They all have “Stratum” equal to 16 instead of the 1 or 2 they should be, and the “When” fields are all blank (just a dash). All the statistics are 0.
...
Under Status > System Logs > NTP, all the log entries are nothing but “Soliciting pool server…” messages.

Re-selecting the WAN interface on the list and restarting the NTP service starts everything up properly and everything works again.

When it's not working, I don't see anything in any of the logs (except for the constant "Soliciting pool server" messages in the NTP log) related to this. Anyone have any suggestions on log settings to try to track this down? I'd really prefer not having NTP listening to the WAN interface if at all possible.

[jahonix]

So did you check an entry as "Prefer"???

[beremonavabi]

@jahonix said in NTP Not Working [SOLVED]:

So did you check an entry as "Prefer"???

Yes. It made no difference.

[Gertjan]

@beremonavabi said in NTP Not Working [SOLVED]:

Was there a specific article you were trying to point me to?

Wow - I guess I checked in after the storm.
Indeed, I was using Google myself to find some possible related posts : ntp not being able to contact remote time servers (Firewall and gateway issues).
Clear is now : there is a something that is not 'default'.
NTP settings : I haven't checked the WAN interface - only my LAN interfaces (I thought this ensures that ntp is serving my local devices with a time server, my pfSense box). See image below.

My Outbound NAT is pretty big and pretty default :

My WAN IP is RFC1918 (192.168.1.10.0/24)
LAN and OPT1 are default.

My NTP settings (tghis time with WAN selected - that works also for me ):

Btw : selecting WAN isn't big deal. As long as there is no firewall rule letting in connections, your ok.
You should know that the GUI is also listening on ALL interfaces (WAN included !).

The only thing I changed there was "fr.pool.ntp.org" (I'm living in France) and checked "this is a pool).

Did you change anything on the Services => NTP => ACLs tab ?

[beremonavabi]

On my WAN interface, I'm allowing IP4 UDP traffic going to port 443 (for OpenVPN clients connecting to my OpenVPN server):

My NTP ACL tab is default: Kiss-o'-death, Modifications, Peer Association, and Trap Service are all checked, while Queries and Service are unchecked.

EDIT: And since so few of us are having this problem, there's got to be something non-default somewhere that's causing this (as you said). I just can't find it.

[beremonavabi]

According to:

http://support.ntp.org/bin/view/Support/TroubleshootingNTP#Section_9.8.

ntpd requires full bidirectional access to the privileged UDP port 123

I assume that's what kpa was referring to, above. So, for those of you who do have NTP synchronizing time correctly without having the WAN interface selected (apparently, most everyone), do you have a firewall rule opening port 123 to UDP traffic on your WAN?

Also, according to:

http://support.ntp.org/bin/view/Support/TroubleshootingNTP#Section_9.9.

there's supposed to be an ntp.conf file at /etc. I don't see that on my SG-4860. Does pfSense use that file for NTP?

[Gertjan]

@beremonavabi said in NTP Not Working [SOLVED]:

According to:

http://support.ntp.org/bin/view/Support/TroubleshootingNTP#Section_9.8.

ntpd requires full bidirectional access to the privileged UDP port 123

Yep.
Part "9.8. Check the NTP port" also states :

Bla bla bla bla ...
If this is not possible, you may need to run ntpd on the firewall itself, so that it can have full unrestricted access to UDP port >123 in both directions, and then have it serve time to your internal clients. However, this may also be disallowed.

Of course, our ntpd is running on the firewall pfSense with 'root' privileges, so it can snap to this "123" port, so it can send out requests , and receives the replies.

I assume that's what kpa was referring to, above. So, for those of you who do have NTP synchronizing time correctly without having the WAN interface selected (apparently, most everyone), do you have a firewall rule opening port 123 to UDP traffic on your WAN?

No way !!!!
As seen above, the ntpd process can go outside anytime it wants to do so - connect itself as a "client" to a ntp "server". These connections are outbound.
Firewall rules on the (WAN) interface are for inbound connection only.
I have no (well, yes : one : an incoming VPN rule) firewall rules on WAN, so by default no initial connection comes in.
NTP servers like fr.pool.ntp.org do not connect to my pfSEnse. It's my pfSense that connects to fr.pool.ntp.org.

Also, according to:
http://support.ntp.org/bin/view/Support/TroubleshootingNTP#Section_9.9.
there's supposed to be an ntp.conf file at /etc. I don't see that on my SG-4860. Does pfSense use that file for NTP?

Yes, no ... this is FreeBSD.
See your post number 4 here ( you !) told us were we (you !) can find this file ^^
As you said yourself over there : the file is here : /var/etc/ntpd.conf

[beremonavabi]

One more attempt at figuring this out.

I saw a bug report that was similar (though it involved CARP):

https://redmine.pfsense.org/issues/5548

The response was:

You're breaking NTP connectivity on the backup by sending the traffic using a CARP IP. It won't, and can't, receive those replies - they go to the primary. When WAN isn't bound, it's probably hitting NAT to a CARP IP because it has a private source IP. NAT it to the WAN IP in that case.

The OP responded:

Added the following NAT rules at the top of the Outbound manual rules list:

Interfc Source Src Pt Dest Dest Pt NAT Addr NAT Pt Static Description 
WAN1 This Fw udp/* * udp/123 WAN1 addr * NO NTP to WAN1 INTFC IP 

and that fixed it.

I was wondering if my restricting all outbound traffic to go through my VPN might be causing a similar problem here. So, I added the following outbound NAT rule (even though it ought to have been covered by my existing NAT rules and I've got "redirect-gateway def1; in my VPN client Custom Options to make sure the firewall, itself, can get out the default interface) which, I think, is an equivalent:

It doesn't make any difference. NTP still won't start without the WAN being selected. Can anyone confirm that these situations might be equivalent and that the NAT rule I added makes some kind of sense?

[stephenw10]

Can we see your full outbound NAT page?

Steve

[beremonavabi]

@stephenw10

[stephenw10]

Hmm, looks OK. And you're using manual mode there or hybrid?

The the WAN the default gateway on your system?

Is default gateway switching enabled?

I could imagine NTP trying to use the VPN gateway and sourcing from something not NAT'd there. It should always use the default gateway though.

Steve

[beremonavabi]

@stephenw10 said in NTP Not Working [SOLVED (mostly)]:

Hmm, looks OK. And you're using manual mode there or hybrid?

The the WAN the default gateway on your system?

Is default gateway switching enabled?

I could imagine NTP trying to use the VPN gateway and sourcing from something not NAT'd there. It should always use the default gateway though.

Steve

Manual mode and, yes, the WAN is the default gateway. Default gateway switching is OFF under System > Advanced > Miscellaneous.

In normal operation, everything is on the VPN_LAN interface (192.168.20.0/24). My firewall rules for that are:

The first two rules are special cases that are hardly ever used. The third rule sends local device traffic to other local devices out the default. And the last rule sends traffic to the outside world out via the VPN's Gateway Group. So, in general, everything to the outside world goes out over the VPN via that Gateway Group. But, I've got "redirect-gateway def1;" in my VPN client Custom Options to make sure the firewall, itself, can still get out the default gateway. I assume NTP falls under that and should get out via the WAN.

That special Outbound NAT rule with "This Firewall", above, was just a desperate stab at trying to make sure that NTP traffic could get out the WAN and not get stuck by the VPN. But, it didn't work. Assuming I wrote that rule correctly, it doesn't look like that's the issue.

I wonder if it could be a DNS issue (not being able to resolve the NTP pool names to addresses -- though why adding the WAN to the NTP listening interfaces would "fix" that, I don't know). I'm using DNS Resolver in NON-Forwarding mode. It's active for all my local LAN-type interfaces (not WAN) and sends everything out via the VPNx_WAN interfaces (again, not WAN). I wonder if I should add the WAN to the Outgoing Network Interfaces:

[beremonavabi]

@beremonavabi
It doesn't look like it's a DNS issue, either. I stuck the actual IP address for a public DNS server

http://support.ntp.org/bin/view/Servers/PublicTimeServer000011

in and removed the WAN interface. Same problem: NTP doesn't start. Put the WAN back in the list and all was well.