one A record to two IP addresses?

al · Guest

Hi all,
I am really not sure where to post this question but I want to make sure
that I ask the experts.
We have a webserver www.website.com (just an example name).
We are going to connect to 2 different ISPs.
My question is, is it possible to point www.website.com to two different IP
addresses?
We just want to make sure that if ISP1 is down, people can still browse our
website via ISP2.
Please give me some suggestion please.
Thanks,
Al

fugi · Guest

In comp.dcom.sys.cisco al <al@somplace.com> wrote:

Walter Roberson · Guest

In article <vht90ul1uqhv61@corp.supernews.com>,
fugi <fugi@ultra.bl.org> wrote:
:when BIND sees 2 IPs for a record, it round-robin's it. it doesn't
:check if it is down before handing out the IP.

BIND certainly does not check to see whether IPs are down before they
are handed out, but your description of round-robin is a bit misleading.

Old versions of BIND did not do any kind of round-robin; 4.9.3 appears
to be the first version that supported round-robin.

When a version of BIND does do round-robin, it is not the case that
it only hands out one of the addresses. The round-robining that
goes on concerns the -order- of the addresses when they are handed out.

See http://www.netsys.com/sunmgr/1997-06/msg00035.html

--
"No one has the right to destroy another person's belief by
demanding empirical evidence." -- Ann Landers

Barry Margolin · Guest

In article <nexTa.1$LS5.238795@newssvr14.news.prodigy.com>,
al <al@somplace.com> wrote:

fugi · Guest

In comp.dcom.sys.cisco Walter Roberson <roberson@ibd.nrc-cnrc.gc.ca> wrote:

Walter Roberson · Guest

In article <vhu06sldn6la7@corp.supernews.com>,
fugi <fugi@ultra.bl.org> wrote:
|In comp.dcom.sys.cisco Walter Roberson <roberson@ibd.nrc-cnrc.gc.ca> wrote:
|> In article <vht90ul1uqhv61@corp.supernews.com>,
|> fugi <fugi@ultra.bl.org> wrote:
|> :when BIND sees 2 IPs for a record, it round-robin's it. it doesn't
|> :check if it is down before handing out the IP.

|> are handed out, but your description of round-robin is a bit misleading.

|what's misleading about it? alternating the order of the records
|returned is the definition of round-robin. I didn't state how
|round-robin worked I simply stated that that was the implementation
|for BIND when more than 1 IP address for the same host is found.

In your second sentance, the leading 'it' is understood to imply
'BIND'. The second 'it' is singular, and so must refer back to a
singular item in the first sentance. The singular item in the first
sentance is 'record', with IPs appearing only in the plural. It makes
no sense, though, to say that BIND checks if a record is down, and in
mentally correcting the second sentance, most people would be led to
the implication that BIND only hands out a singular IP, rather than to
the mental correction that the second sentance must apply distributed
across each of the IPs in the record.

The second singlular 'it' in the first sentance refers to 'record' to
preserve number. Saying that the record was round-robin'd (instead of
the IP addresses within the record) is a bit unusual but not invalid,
but in being unusual again most people would be led to the
understanding that you are saying that BIND will only hand out one IP
at a time and will do so in round-robin order.
--
Oh, to be a Blobel!

CybrSage · Guest

Reading the post by Fugi, I understood it to say:
"BIND does not check to see if an entry is working. When BIND sees two
entries, it hands out the first one, then uses round-robin to move that
entry to the end of the list, and hands out the next entry when it receives
a request."

From Walter's entry below, I have become lost in his plural,singular, it,
its...what Walte forgot is that Fugi was actually answering Al, the original
poster. His use of the word it (check if IT is down) refers to the original
poster's question.

The statement Walter makes about "It makes no sense..mentally correcting" is
unsound. The original poster (Al) does (or at least did) not know how BIND
or DNS works, and was asking. He cannot do any "Mental Corrections".
Remember, the post Fugi made was in reply to Al, and should be read as such.

"Walter Roberson" <roberson@ibd.nrc-cnrc.gc.ca> wrote in message
news:bfn1vg$jhu$1@canopus.cc.umanitoba.ca...

Walter Roberson · Guest

In article <p8CcnajKpcIioIKiRTvUqA@comcast.com>,
CybrSage <CybrSage13@Hotmail.com> wrote:
:Reading the post by Fugi, I understood it to say:
:"BIND does not check to see if an entry is working. When BIND sees two
:entries, it hands out the first one, then uses round-robin to move that
:entry to the end of the list, and hands out the next entry when it receives
:a request."

Which is not what it does: it hands out all the IP addresses, but it
changes the -order- it lists them in.

In other words, you -were- misled by Fugi's posting. If you analyze
the sentances to see what parts of his phrasing led you to the
misunderstanding, I believe you will find that the conflict between
singular and plural bore a substantial role.
--
IMT made the sky
Fall.

al · Guest

Is it possible to setup priorities to the IP addresses meaning I prefer
routing via ISP1 and only route it via ISP2 if ISP1 is down.
Just like setting up a secondary MX record.
Thanks for your reply,
Al

--

-------------------------------------------------------------------------
FIGHT BACK AGAINST SPAM!
Download Spam Inspector, the Award Winning Anti-Spam Filter
http://mail.giantcompany.com

"fugi" <fugi@ultra.bl.org> wrote in message
news:vht90ul1uqhv61@corp.supernews.com...

Walter Roberson · Guest

In article <6%HTa.669$576.466@newssvr24.news.prodigy.com>,
al <allen@somplace.com> wrote:
:Is it possible to setup priorities to the IP addresses meaning I prefer
:routing via ISP1 and only route it via ISP2 if ISP1 is down.
:Just like setting up a secondary MX record.

Not really, no.

BIND's round-robin behaviour is an option; if it is not turned on,
then the IP addresses will be handed out in the order configured.

*Most* clients will try the addresses in the order they are returned
in, but it is considered acceptable for clients to use whichever
address they prefer, such as by examining each of the addresses and
chosing the address with the fastest route. The order is thus a
hint to many sites, but other sites will do whatever they feel like.

If you wanted to change the system to only hand out the address
on ISP1 as long at the link was up, then you will face problems
with caching. Your expiry period on the records would have to be no
longer than you were willing to have a remote site continue to
make new connections via ISP2 when ISP1 is back up. For example,
if you wanted sites to start using the ISP1 address again within
(say) 5 minutes, then your cache expiry period would have to be
5 minutes -- but anything less than 1 hour is considered bad
Internet manners.

There is also the issue hinted at in the above paragraph: that the
address handed out for DNS purposes only controls the way that
a connection is *started*. Once a connection is -started- on ISP2,
it is going to continue as long as it takes; e.g., if you have
a "push" WWW page, or you allow connections to be held open
for further requests, then the connection could be up for several
hours or days.

I suspect that what you really should be looking at is arranging
for ISP2 to -route- a single IP address when the link to ISP1 goes
down. That requires setting up a routing protocol such as BGP4,
and not all ISPs are willing to cooperate on that until you
get into the dedicated fibre or T1 price range (i.e., several
hundreds of dollars a month.)
--
Cannot open .signature: Permission denied

Bit Twister · Guest

On 24 Jul 2003 03:31:16 GMT, Walter Roberson wrote:

Walter Roberson · Guest

In article <slrnbhul8b.91n.BitTwister@wb.home>,
Bit Twister <BitTwister@localhost.localdomain> wrote:
:On 24 Jul 2003 03:31:16 GMT, Walter Roberson wrote:

:> I suspect that what you really should be looking at is arranging
:> for ISP2 to -route- a single IP address when the link to ISP1 goes
:> down.

:I was wondering when we would get down to this point.

:When I enter the OP's url, my ISP (ISP0) would have to get an ip addy
:resolved to which ip address (isp1 or isp2)?

:When that isp is down, I still would hit 401.

In the BGP scenario, there would only -be- a single IP address, but
it would be an IP address which was not permanently assigned to
either ISP1 or ISP2 -- a "portable" address. ISP1 would -route-
the IP address to the OP over top of a carrier /29 or /30 that -is-
in ISP1's address space. ISP2 would also -route- the same IP
address to the OP overtop of a /29 or /30 part of ISP2's address space.
However, the routing priorities would be set up so that ISP2 would
surpress advertising the route to the outside world as long as
the ISP1 link was up.

Within a short time after the ISP1 link went down, ISP2 would
be notified to start advertising the route to the OP's public
IP address. The routing change would then propogate out to the
other networks, until eventually a router somewhere
downstream from your ISP0 that used to make the path decision
noticed the change and started using the new path.

I might have the details of which system does the advertising wrong.
I've never configured BGP, just read enough about it to know its
approximate function.
--
Oh, yeah, an African swallow maybe, but not a European swallow.
That's my point.

Freeride · Guest

On Wed, 23 Jul 2003 14:47:47 +0000, al wrote:

fugi · Guest

In comp.dcom.sys.cisco Walter Roberson <roberson@ibd.nrc-cnrc.gc.ca> wrote:

Brooks Hagenow · Guest

"Walter Roberson" <roberson@ibd.nrc-cnrc.gc.ca> wrote in message
news:bfnhhs$q50$1@canopus.cc.umanitoba.ca...