Testing SRV records in DNS for SIP with CVP 4.0

geoff · ‎06-20-2008

I have configured the SRV records for two Proxy servers in DNS. It's all working and when the primary CUPS is down, the secondary is used. When both are up, I don't think my priorities are working.

I have both udp and tcp records configured and they specify two A records - the records have equal weight, but one has priority 1 and the other priority 2. I have a strong preference for one CUPS over the other, so this is the way I want this to be, rather than load balancing with equal priorities.

If I call nslookup from a command prompt and type 'set type=SRV' I can query the cluster name OK.

All the details come back correctly, including the weights and priorities. For example:

C:\installed>nslookup

Default Server: us-hqtul-adhp01.ucce.company.icm

Address: 10.10.105.11

> set type=SRV

> _sip._udp.cups.ucce.company.icm

Server: us-hqtul-adhp01.ucce.company.icm

Address: 10.10.105.11

_sip._udp.cups.ucce.company.icm SRV service location:

priority = 2

weight = 10

port = 5060

svr hostname = us-drtul-prhp01.ucce.company.icm

_sip._udp.cups.ucce.company.icm SRV service location:

priority = 1

weight = 10

port = 5060

svr hostname = us-hqtul-prhp01.ucce.company.icm

us-drtul-prhp01.ucce.company.icm internet address = 10.10.202.12

us-hqtul-prhp01.ucce.company.icm internet address = 10.10.102.11

> _sip._udp.cups.ucce.company.icm

Server: us-hqtul-adhp01.ucce.company.icm

Address: 10.10.105.11

_sip._udp.cups.ucce.company.icm SRV service location:

priority = 1

weight = 10

port = 5060

svr hostname = us-hqtul-prhp01.ucce.company.icm

_sip._udp.cups.ucce.company.icm SRV service location:

priority = 2

weight = 10

port = 5060

svr hostname = us-drtul-prhp01.ucce.company.icm

us-hqtul-prhp01.ucce.company.icm internet address = 10.10.102.11

us-drtul-prhp01.ucce.company.icm internet address = 10.10.202.12

> exit

When I do this twice, as shown above, the order changes. But the priorities are such that us-hqtul-prhp01.ucce.company.icm should be first.

I get the same "flipping" of the order if I use the DNS utility from the CVP diagnostic template.

My question is this: is nslookup the way to test this? Should the order oscillate? Is testing out of the diag servlet subject to the same rules?

Regards,

Geoff

pantinor · ‎06-20-2008

no the oscillation is not normal.

Is this Microsoft DNS server or linux based (bind?) I have never seen this before.

nslookup is the tool to use.

Is it possible that different DNS Servers are being used, if you have several of them in your TCP configuration for DNS.

One thing to try is run the sniffer like wireshark with the "dns" display filter turned on. it may show something you are not aware of regarding which DNS server you are contacting.

Can you attach a text file that shows your DNS configuration? ie in linux it is a "zone" file.

Another thing to try is find other examples of SRV configurations for your DNS server and stare/compare. There may be something special about the way the records are configured.

geoff · ‎06-20-2008

Good to get a reply. Thanks.

This is Microsoft DNS. There are two DNS servers, Active Directory Domain Controllers for the ICM domain. So the DNS are sync'd.

I am using the two DNS from the client I am running nslookup from. Just a second -- I'll go to just one. Nope, still flips.

The only pictures I have is in the Config guide, and it looks similar.

Are you saying that with a set up like I have, with A record at priority 1 and the other A record at priority 2, nslookup (and the DNS lookup from CVP diag) should always return the records in the same order - the high priority one first?

Regards,

Geoff

adignan · ‎06-22-2008

Hey Geoff,

What I "believe" you want to do is set the priorities to be the same and change the weight to be 100% for one and 0% for the other.

_sip._udp.cups.ucce.company.icm SRV service location:

priority = 1

weight = 100

port = 5060

svr hostname = us-hqtul-prhp01.ucce.company.icm

_sip._udp.cups.ucce.company.icm SRV service location:

priority = 1

weight = 0

port = 5060

Let me know if that works (it should).

adignan - CDW Berbee

*please rate helpful posts

Chad Stachowicz · ‎06-22-2008

Oddly enough, I had some time on my hands this morning, and I have been meaning to configure this in our lab for testing, so I went to try. I can send calls to CUPS6.demo.com (10.1.1.16) and CUPS6-SUB.demo.com (10.1.1.17) by both IP and DNS hostname. However when I configure an SRV records containing these 2 hosts, my call always fails. Is there anything special that needs to be done on either CVP or CUPS to enable SRV functionality? I tried configure Microsoft DNS a few different ways. But _sip._udp.demo.com

port 5060

pri 1

weight 100

host CUPS6.demo.com

_sip._udp.demo.com

port 5060

pri 1

weight 0

host CUPS6-SUB.demo.com

Haven't gotten a call to succeed through it yet. Also I tried using nslookup and everything looks ok. Any suggestions?

Chad

Chad Stachowicz · ‎06-22-2008

When I run nslookup I get

Server:UnKnown ?

This this could be messing things up?

Chad Stachowicz · ‎06-22-2008

Well,

I got DNS SRV working for the primary CUPS proxy. I am running into something strange. My secondary proxy will accept calls into it via IP ONLY when the primary is online. If I shut off the primary, then the calls into the secondary immediatly start failing. CUPS 6.0.1.1000-21... I believe that my DNS SRV is correct now, I just can't test since the backp server won't process calls when the primary is down.

Thanks for the guidance!

Chad

Chad Stachowicz · ‎06-22-2008

Andy / Geoff,

I see this, and Geoff I don't get accurate tests the way you were doing it, I'm also looking for a good way to test this because I want my routers, and call servers to use there respective Sip Proxy and the way for me to do this yet achieve redundancy is through DNS SRV. If I find a good answer I will let you know, is there maybe a debug to watch packets leaving the router?

Chad

Chad Stachowicz · ‎06-22-2008

From my testing and reading today nslookup is not a valid way of testing priority. Anyways if the priorities are different aka

pri 1 and pri 2

pri 1 will always be chosen if its up!

if they are the same at pri 1 weight 50 and pri 1 weight 50

it will be perfectly load balanced 50 50

if its configured like andry has it the one with weight 100 will always be chosen, same as setting it to priority 1 and priority 2.

HTH,

Chad

geoff · ‎06-27-2008

This is dead easy to test. All you need is a DNS, no CVP needed. I made this test on a Domain Controller while I was idling away my time on the booth at Cisco Live. Quickly added a domain to the forward lookup zone, added SRV records for TCP and UDP with two host names (A records) that already existed. Ran nslookup. Same result.

The specs on SRV say that weights are only examined if priorities are the same. Anyway, nslookup on Windows always alternates the answer.

Still looking for feedback on this.

Regards,

Geoff

Chad Stachowicz · ‎06-27-2008

Geoff, I don't think thats the proper way to test, I think doing a packet capture during a phone call redirecting to see which SIP Proxy is hit is the way to test, I can give this a run through on monday!

Chad

geoff · ‎06-28-2008

I agree that something seems odd with either nslookup or Microsoft DNS. Wireshark is probably a useful strategy. I can tell from the logs on the SIP Proxies which one is being hit. I'll also do some testing on Monday. Cheers.

(PS. I met a couple of your Touchbase mates at Cisco Live.)

Regards,

Geoff

Chad Stachowicz · ‎11-02-2008

Geoff,

I never saw this, you must have met corey and Martin Blake :) Good guys, they flew to The San Fran office after Cisco Live, and I met the for the first time. Also David came from our ofice and was down there. Hopefully next year I will go myself!

Cheers!

Chad

geoff · ‎11-02-2008

Mate, I know Corey from way back when we worked at HP in Brisbane, Oz building Smart Contact. Corey knows a ton of stuff - Touchbase are fortunate to get him. ;-)

Regards,

Geoff