Página de Inicio › Forums › HAfs (High Availability for FreeSWITCH) › Configuration & Upgrade › Tutorial: SIP using NAPTR and SRV DNS Records
Tagged: SIP TCP NAPTR SRV
-
AuthorPosts
-
Telium’s High Availability for Asterisk (HAast) and High Availability for FreeSWITCH (HAfs) clustering software can use a variety of methods to allow upstream and downstream devices to find the active node (and SIP stack). Use of DNS records in one such method, and the basic principles are described below.
SIP can be run, for example, over UDP, TCP or TCP with TLS (SSL) encryption however standard “A” DNS record lookups won’t tell you anything about which of these protocols to use. NAPTR records are commonly used with SIP in conjunction with SRV records to discover what types of service are available for a name, (such as SIP, email or web) what name to use for an SRV lookup and (using the SRV record) what port and “A” records to use to find the IP for the service. That might come off sounding somewhat complicated so lets take a look at the entire process through an example.
Lets consider a call to 2125551212@example.com. Given only this address though, we don’t know what IP address, port or protocol to send this call to. We don’t even know if example.com supports SIP or some other VoIP protocol like H.323 or IAX2. I’m implying that we’re interested in placing a call to this URL but if no VoIP service is supported, we could just as easily fall back to emailing this user instead. To find out, we start with a NAPTR record lookup for the domain we were given:
# host -t NAPTR example.com
example.com NAPTR 10 100 “S” “SIP+D2U” “” _sip._udp.example.com.
example.com NAPTR 20 100 “S” “SIP+D2T” “” _sip._tcp.example.com.
example.com NAPTR 30 100 “S” “E2U+email” “!^.*$!mailto:info@example.com!i” _sip._tcp.example.com.Here we find that example.com gives us three ways to contact example.com, the first of which is “SIP+D2U” which would imply SIP over UDP at _sip._udp.example.com. But first, a bit about all the fields we have here.
In the first line of our result in the example above, example.com is the name we looked up and NAPTR is obviously the type of record. The 10 refers to the preference for the record. The lower number is always tried first. 100 is the order and is only important if the preference numbers are the same.
The “flag” field, in this case “S”, is next. There are currently four possible flags: “S” which denotes that an SRV lookup is to be performed on the output of this NAPTR record. “A” means the result should be looked-up as an “A”, “AAAA” or “A6” record. A “U” means that the NAPTR result is an absolute URI that the application should process. A “P” would signify a “non-terminal” rule where additional NAPTR lookups would be necessary. It is application specific and can be mutated by regular expressions. (discussed below)
Next we have the “services” field, “SIP+D2U”, “SIP+D2T” and “E2U+email” in the example above. “SIP+D2U” is SIP over UDP, “SIP+D2T” is SIP over TCP and (you guessed it) “E2U+email” stands for email. This is the application specific service options we have to reach example.com.
It might be hard to notice the next field, “”, because there is nothing there, but this is the “regular expression” field. The regular expression is used to mutate the original request (in this case “example.com”) into something new. We’re not using it here but you could use this to substitute the entire name or parts of the name used in the original query. (NOTE: These are NOT cumulative. You would never use a regular expression on the output of a NAPTR lookup, only on the original query.)
The last item we have is the “replacement”. In the first result from our example above, we have “_sip._udp.example.com”. Regular expressions and replacements are mutually exclusive. If you have one, you shouldn’t have the other. The replacement is used as the “result” of the NAPTR lookup instead of mutating the original request as the regular expression in the paragraph above.
That’s all of our fields. So because we have the “S” designation in the “flag” field, our next step is to find the SRV record for the replacement “_sip._udp.example.com”.
# host -t SRV _sip._udp.example.com
_sip._udp.example.com SRV 5 100 5060 sip-udp01.example.com.
_sip._udp.example.com SRV 10 100 5060 sip-udp02.example.com.We get two answers here so first we’ll try sip-udp01.example.com because it has the lower of the two priorities. (priority 5 before priority 10. 100 is the weight which is used to differentiate between records of the same priority.) Next we do an “A” record lookup to find the IP of the server to use to send our SIP INVITE.
# host sip-udp01.example.com
sip-udp01.example.com has address 11.22.33.44So in this example, our top preference would be to send a SIP INVITE via UDP to port 5060 on 11.22.33.44. Failing that, we would look up the IP for the other SRV response (sip-udp02.example.com) and hit that via UDP on port 5060 as well. Failing all of that, we would go back to the next response we got via the original NAPTR lookup and do an SRV lookup on _sip._tcp.example.com and presumably try a TCP connection to some other server and port combination. And lastly, failing all of that, the last response from the NAPTR lookup has us sending an email to info@example.com.
Of course this usually isn’t done on the command line, but by an application. It is handy, however, to see how you can mimic the requests a VoIP application is going to make for illustration and troubleshooting purposes.
Not all clients will be able to speak all protocols so you should try to supply some alternate methods of contact in your NAPTR response rather than just one protocol in practical implementations. This could become particularly interesting in a fully IP world when, for example, my “contact info” is anders@example.com. The NAPTR record would return several ways for me to be contacted perhaps via VoIP, an IM option and lastly an email option. Remember, VoIP calls aren’t restricted to numbers only, so as long as a client supports it, NAPTR allows for your email address to also be your VoIP “number”.
While most major DNS server packages out there today support NAPTR and SRV records natively, some do not. In the case of djbdns’s tinydns authoritative nameserver, there is a patch for NAPTR support but you can also describe NAPTR records in the generic record format. There is a djbdns NAPTR record builder that creates NAPTR records in the generic syntax for use with an unpatched djbdns tinydns server.
Note: Be aware that some applications (such as OpenSER / OpenSIPS) prepend the protocol information (_sip._tcp or _sip._udp) to names automatically before doing the SRV lookups. Check your nameserver log to see what clients are asking for.
Anders Brownsworth
It’s important to understand that the RFC which specifies how NAPTR/SRV records are implemented is rather vague in regards to how they are used (see https://datatracker.ietf.org/doc/html/rfc2782). The algorithm which outlines how the UA should perform the lookup (see page 7) does not specify when the priority list is created, nor when the UA should restart at the top of the priority list. The result of this ambiguity is very different behaviors on the part of UA clients in regards to how they respond to a change in availability of a SIP server. To make matters worse, each different implementation may in fact be fully compliant with the RFC (and manufacturers will all claim their way is the right way).
From a practical standpoint, some UA clients do an excellent job participating in the failover of the active SIP node, while others do not. For example, SNOM phones (as of firmware released in 2013) select the SRV record to use in priority order, but starting at the current priority. The result is that SNOM phones are immediately responsive, and in the event of a failover they quickly detect the active SIP stack and future calls/registrations are once again immediately responsive. SNOM is fully complaint with RFC 2782.
In comparison, Panasonic phones will iterate through the list of SRV records in order of priority upon every registration/call. Every time they do so they start at the lowest priority and move upwards. So unless the first SRV record is currently in use, it can take 30-45 seconds per SRV record to find the active SIP stack, for every call/registration. Companies using Panasonic phones discover that in the case of a failover it takes 30 sec-4 min per call for the phone to connect to the SIP server. As noted above, Panasonic is fully compliant with RFC 2782.
As manufacturers get more real-world HA experience they tend to use SRV records differently. For example, manufacturers like Cisco changed their SRV lookup behavior to be more like SNOM. As of firmware 8.5(3), Cisco 7941/61, 7942/62, 7945/65, and 7906 phones stop attempting to reach the SIP hosts in priority order starting at the lowest on every call/registration, they now (as of that firmware release) check SRV records starting with last priority used. For this reason selected Cisco and SNOM phones work very well in failover scenarios.
If you are designing a high availability telephony environment you must consider the behavior of the UA clients. If you control the make/model/firmware/configuration of each UA client, then you may be able to use SRV records for active SIP node contact. And, NAPTR/SRV records are the preferred method of locating the active SIP node. However, if you can’t control the UA clients in use or the UA client chosen has a poor implementation of NAPTR/SRV records then you should switch to allowing HAast to control the location of the active node. This can be done through HAast event handlers updating DNS records, changing routes, modify firewall configurations, etc.
Telium has implemented a wide variety of solutions across a large range of hardware/software platforms. We would be pleased to design a solution which fits your needs, and to implement it as well. Please note that we cannot rate, recommend, or discourage use of a particular phone / UA client for legal reasons. The comments above reflect our experience with particular phones at a point in time, and these comments do not constitute and endorsement of any particular product nor a criticism of any other product. Since each phone may be fully RFC 2782 compliant we are not saying that any manufacturer’s implementation is wrong or poor, rather we are saying that certain phones / UA’s work better in certain scenarios.
-
AuthorPosts
- You must be logged in to reply to this topic.