quarantined email release fails

Hi Adrie,

No issue reported yet. Check in the smtp.log when you release the quarantined mail, do you see any errors? 

"Releasing has recently gone wrong on my macos Sierra machine." Did you mean that the emails are releasing perfectly through a Windows system?

Thanks

I wouldn't know, I do not have access to a windows PC. 

What I know is that, on the same macos machine, Firefox and Chrome are working.

Cheers. Adrie

testing the fqdn with openssl

prosserg@ITRoom-Mint ~ $ openssl s_client -connect utm.<fqdn>:3840
CONNECTED(00000003)
140409629832864:error:140770FC:SSL routines:SSL23_GET_SERVER_HELLO:unknown protocol:s23_clnt.c:795:
---
no peer certificate available
---
No client certificate CA names sent
---
SSL handshake has read 7 bytes and written 295 bytes
---
New, (NONE), Cipher is (NONE)
Secure Renegotiation IS NOT supported
Compression: NONE
Expansion: NONE

The utm has no virtual server configured for SSL on port 3840, yet the utm is causing the browser (I'm now using Firefox on Linux) to switch from http to https

BUT as previously noted, the same release link works fine on an iPhone (Safari)

Gary

Of course that is what I was doing and reporting, from a PC inside the network and from a PC outside the network. I was confirming that DNS is correct.

Thanks for the tip, but...

Please explain what you mean by "Probably I know the answer." What IS the problem and why is this an answer ?

Gary

Test the Host configuration as I said before.
You are entering in loopback. And you didnt mentioned your IPHONE Connection 3G or wifi

If You are from outside you can release the email from your external IP, from inside from your internal
Hope you are clear on tha

Oldeda

iPhone works fine on wifi and on 3G

utm httpd: 192.168.15.138 - - [27/Mar/2018:13:24:26 +0100] "GET /release.plc? <----internal device

utm httpd: 85.255.237.84 - - [27/Mar/2018:13:25:19 +0100] "GET /release.plc? <----external device

You say 'you are entering in loopback' - I appreciate you are trying to help, but that phrase doesn't make sense to me. 

You say test this configuration

Name:  utm
Type:   Host 
Ip4 Address: 192.168.2.1 (your internal interface IP) 
DHCP: No DHCP Server
DNS Settings: utm.hidden-domain.com (your FQDN that is equal in Quarantine Configuration Hostname)

so what definition do you propose for my other internal networks (5 in total) ?

G

Internal Network Address 192.168.15.138 for all 5 internals, because you can reach this IP from all internal networks. 

Oldeda

can you explain exactly what you meant by 'a loopback' effect ?

I still don't understand why a non-ssl client request to the utm on port 3840 is converted to ssl (on the same port) for SOME but not all clients, thus generating the error I've reported.

Gary

Because you may have some NAT rules for WAN interface, waf or web filter. You are hitting wan address i think

Maybe

but on click release link

web filter - nothing in its log

waf - nothing reported in its log

nat rules - we have 2 which relate to the external IP matching the fqdn I have cited

there is nothing in the web server log of GEORGE_SECURED indicating 'release' requests are being directed there - not surprising since HTTP = port 80, while release request is for port 3840

Gary

Gary, I don't understand - are clients in the same subnet treated differently?  How do you know that some aren't switched from HTTP to HTTPS - is it possible that they're using a different browser or ???

Are you using split DNS so that the FQDN resolves to the Internal IP inside you LANs and to the External IP outside?  If you're not, I would replace those two NAT rules with:

1. DNAT : Internet -> {Group of HTTP & HTTPS} -> External (WAN) [VPN_MyPortal] (Address) : to GEORGE_SECURED
2. Full NAT : Any IPv4 -> {Group of HTTP & HTTPS} -> External (WAN) [VPN_MyPortal] (Address) : to GEORGE_SECURED

Of course, if you have only one LAN on Internal, I would use "Internal (Network)" instead of "Any IPv4" in the second NAT rule.  Also, check out #5 in Rulz.

Cheers - Bob

Gary, I don't understand - are clients in the same subnet treated differently?  How do you know that some aren't switched from HTTP to HTTPS - is it possible that they're using a different browser or ???

Are you using split DNS so that the FQDN resolves to the Internal IP inside you LANs and to the External IP outside?  If you're not, I would replace those two NAT rules with:

1. DNAT : Internet -> {Group of HTTP & HTTPS} -> External (WAN) [VPN_MyPortal] (Address) : to GEORGE_SECURED
2. Full NAT : Any IPv4 -> {Group of HTTP & HTTPS} -> External (WAN) [VPN_MyPortal] (Address) : to GEORGE_SECURED

Of course, if you have only one LAN on Internal, I would use "Internal (Network)" instead of "Any IPv4" in the second NAT rule.  Also, check out #5 in Rulz.

Cheers - Bob

Bob - I'm also confused

it seems none of the available utm logs are recording what's going on, so we have no definitive report

from what I've seen a truly external client can release 

I think it must be some kind of 'trombone' effect - all internal clients see the fqdn as a public IP....

The fact that internal clients get an SSL error is also confusing since AFAIK none of our waf, dnat, firewall or web filtering rules are configured to redirect to ssl. In fact none of our rules cite port 3840  

Gary

That's an indication that replacing the two NAT rules with the ones I've suggested might resolve your issue.  The alternative is split DNS as described in Accessing Internal or DMZ Webserver from Internal Network.

I suspect the SSL error is a byproduct of the real problem.

Cheers - Bob

Bob

in attempting to set up the NAT rule #1 you suggest

we do use split dns although in fact its superfluous since all dns sources resolve the relevant fqdn to the same IP (an external interface of the utm)

Bob

but we are dealing with port 80 and 443 requests in our dnat rules

meanwhile the release request is for port 3840

why is that being ignored by the utm - firewall, waf, web filtering whatever ?

G

You're right that the release is done on a different port, so the repaired NAT rule doesn't make a difference.  See #5 in Rulz to understand why you got the error message.

If you have DNAT and WAF, see #2 in Rulz to understand that the DNAT causes traffic to bypass WAF (reverse proxy).  See the link in my previous post to understand when you need Full NAT instead of a DNAT.

Oldeda is right - the best solution is split DNS where resolution inside your LAN is to the local IP, not the external address where you have nothing to handle port 3840 traffic from inside your LAN.

Cheers - Bob

Bob

I get your advice. And maybe I have to swallow it and move all our dns to the utm via interface based host definitions.

Meanwhile our current subnets, dhcp, dns config works fine considering we have a single utm and so for resilience we use ISP dns (via forwarders) for some subnets like visitor wireless where no internal hosts should be available, or lan based dns servers for subnet specific services, or in two cases an AD domain.

Its currently a config that works for all except, and then only recently, in one case (and only for some internal clients) - ie the spam release.

The bit of angst I feel is that the ssl error I've reported seems to come from the utm itself simply because either

- it doesn't have an ssl capable listener on port 3840

- the utm is redirecting non-ssl requests to ssl; I note you say this is a 'side effect' of NAT rules - really ? Actually we can't tell as there is no relevant logging (AFAIK).

Since the default hostname for spam release is the utm itself (with an ssl cert defined, either self signed or uploaded) why is the utm release url coded as http not https ? 

Gary

This morning I changed the dns record* for utm.<fqdn> from its external interface to my PC's relevant subnet interface

* on our local bind server

Then, from my pc on that subnet, I ping / dig utm.<fqdn> which comes back with the IP of the internal interface.

I try release link - get same ssl error as previously stated. The utm http daemon log has

A little later I also found this in the same log