Hi all,
I couldn't really find a general questions regarding DNS but I know having managed lots of active directory that there's a bit of DNS involved. Having learned quite a bit about DNS originally working with active directory I thought I might see if there's anybody who has dealt with such a class of record before here. Feel free to set me in the right direction if this is not the place to start.
I was wondering if there's any proof or any reason I should believe that web browsers like chrome or firefox support http srv records? this guy seems to wanna indicate that to me: http://www.pantz.org/software/bind/srvdnsrecords.html
One might immediately want to assume that because since that guy started his admin career back in 1999 that it must be true but considering what led me to these findings is the fact that I thought you could just have two A records and the browser would always try one and go to the next if there's more than one but that's not the case (in Chrome anyway.)
I've been told by a handful of people that SRV records are "hard" and that they don't know anything about building a failover cluster of load balancers any other way than with anycast addresses which is completely out of the question for me. I'd be lucky to ever get an AS # let alone a gun with my history and not to mention its totally outside of my budget (apparently anycast requires at least a /24.)
Google groups goes on to fight over whether or not they should add another feature that people don't have the ability to turn on or off arbitrarily: http://www.quora.com/Load-Balancing/Why-dont-browsers-support-SRV-records-for-HTTP
It is clear to me that they are talking about /the/ RFC 2782: http://www.anta.net/nic/draft-andrews-http-srv-01.shtml
And before I drew many of these conclusions I went on to try and see if I could get it to work first, with the intent of confirming the solution afterwards. I'm left with the possibility of something very misunderstood works perfectly because I found that when I added the two SRV records:
_http._tcp.mydomain.com. IN SRV 0 5 80 lb-1-1-1.mydomain.com. _http._tcp.mydomain.com. IN SRV 0 10 80 lb-1-1-2.mydomain.com.
that the errors that I was getting in Chrome that I'm certain were related to the circular dependency between the two load balancers caused by the redirect from HTTP to HTTPS that they respond with stopped happening. It's gotta be a coincidence and there's something else going on and I just need to go with the proverbial stfu and buy lots of IP addresses when I only need two for anycast, right? I'm not really sure how to find out for sure due to the level of complexity in these sorts of stacks.. I'm limited to strace, rndc querylog, ethereal.. do I even need to go that far to find out and am I the only person who doesn't think this should even be an issue?