006 8:38 AM
> To: Declude.JunkMail@declude.com
> Subject: Re: [Declude.JunkMail] OT: Poor man's high reliability?
>
> Side ntoe:
>
> UltraDNS has been having some issues the last couple of days
> due to the blue security ordeal ->
> http://it.slashdot.org/it/06/05/18
Side ntoe:
UltraDNS has been having some issues the last couple of days due to the blue
security ordeal -> http://it.slashdot.org/it/06/05/18/2158227.shtml
Darrell
---
Check out http://www.invariantsystems.com for utilities for Declude, Imail,
mxGuard,
Another option to consider ... www.ultradns.com has a service that does
this. I've never priced it, so it may be pricey.
Rob
---
[This E-mail scanned for viruses by Declude Virus]
---
This E-mail came from the Declude.JunkMail mailing list. To
unsubscribe, just send an E-mail to [EMAIL PROTE
For example, database-backed sites whose database can be open for
writing at only one site are a *helluva* lot harder to balance.
You're right about that. There is some interactivity on the site, but it all
results in emails to the office, so no db sync'ing issues here.
-d
---
This
> ...and to make things a bit more confusing...an NS query to my
> various servers for different domains always sends the first
> response in the registrar order and then it randomizes after the
> first request. So this means that the load should be heavier on the
> primary name
> I think the trick may be whether or not the DNS server that handles
> the client requests round-robins the cache. It appears that Windows
> 2003 DNS does do this, and BIND also appears to do this based on
> tests that I just did.
All modern recursors use an intelligent response-time + roun
> It's the root server's order and the querying server's handling that
> matter.
Absolutely not. It is ONLY the querying recursor's handling that
matters, for it is that recursor that is talking to the authoritative
NSs, and it makes a more "locally intelligent" choice than you think.
> ..
> I only raise the issues about primary and secondary because all my
> domains have dns.skywaves.net as the primary. That is a deicated
> name server on a DS3, and it is never remotely overloaded. But
> dns.skywaves.com, on a separate line at home, gets an awful lot of
> inquiries.
Y
...and to make things a bit more confusing...an NS query to my various
servers for different domains always sends the first response in the
registrar order and then it randomizes after the first request. So this
means that the load should be heavier on the primary name server as
registered, bu
Dave,
I think the trick may be whether or not the DNS server that handles the
client requests round-robins the cache. It appears that Windows 2003
DNS does do this, and BIND also appears to do this based on tests that I
just did.
So maybe it does spread things evenly. I don't operate my ow
> The registrars aren't returning anything to non-authoritative
> recursors, the roots are.
[Well, slight correction to this, if the registrar is _also_ hosting
DNS, then they are of course part of the treewalk. But the two
functions are not linked, regardless of what NetSol
Hi Sandy -
And that answers the question I just posed to Matt regarding my two name
servers.
Thanks.
-d
---
This E-mail came from the Declude.JunkMail mailing list. To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type "unsubscribe Declude.JunkMail". The archives can be fo
Hi Matt-
In any event, the root servers will return a list of name servers. If the
first name server returned is offline, then the DNS client should try the
second, regardless of which is considered "primary" and which is "secondary"
as far as the registrar is concerned.
I only raise the iss
> Maybe I'm missing something, but if you query one of the root
> servers for a domain's NS servers, they are always returned in the
> order that they are listed in your domain name registration.
"Registrar order" is not observed by the recursors querying the
authoritative namese
> This actually happened to me recently: The potential customer, who
> is now hosting with some $10/month outfit, thinks that $50/month is
> about all he can muster for a redundant solution. And you have to
> agree with him, sort of. A 5X increase in his hosting bill seems
> like an awfu
> I've experienced it both ways. It seems that some registrars return
> the DNS servers primary-first, but NetSol at least, in my
> experience, returns a name server in random order.
The registrars aren't returning anything to non-authoritative
recursors, the roots are. And th
Hi Jay-
You are absolutely right, but the customer doesn't see it that way. If he
can get one website hosted for $10/month, he ought to be able to get two
hosted for $15, and if we screw up and lose some orders, there's another
$10/month outfit out there somewhere to replace us.
Ordinarily,
It depends on what you query. If you query one of the root servers,
even netsol.com's NS records always appear in order, but if you query a
server like ns1.netsol.com, it will round robin the order. It's the
root server's order and the querying server's handling that matter.
My understanding
thinking about first, not the monthly cost of the hosting.
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Dave Doherty
Sent: Thursday, May 18, 2006 11:50 PM
To: Declude.JunkMail@declude.com
Subject: Re: [Declude.JunkMail] OT: Poor man's high reliab
I've experienced it both ways. It seems that some registrars return the DNS
servers primary-first, but NetSol at least, in my experience, returns a name
server in random order.
And I think that the root servers follow the lead of the registrars, no?
-d
---
This E-mail came from the Declud
Er, no... that isn't the way NS records are used. Recursors looking up
your NS records do not "know" that you went to Dotster or whatever and
entered ns1 before ns2. The NS records handed out by the roots are
used by recursors in what can be called a random order (the order
actually ha
Hi Matt-
Agreed on all points.
I'm not looking for perfection or load balancing. The idea is to provide a
low-cost response to the question "So what happens to my site when the disaster> occurs?" and the customer deos not want to pay the freight for
a real solution.
This actually happened to
> It seems to me that this would work, but I've never tried it. What
> do you DNS gurus think about the idea?
In addition to my thoughts in the other post, I'd say that there would
be no problem with such an approach, as long as you are absolutely
sure that the fate of the HTTP daemon
> So it wouldn't be perfect, but it would probably work decently as a
> fail-over.
Considering that there's no other way other than short TTLs to do
geographic load-balancing w/failover -- even mega-hitting GLB hardware
has a rather explicit reliance on the TTLs it gives out based on
r
The only issue here is that many systems will cache beyond what you set
things at. The standard cache time for an A record is only one hour,
but large ISP's will override the cache settings and set the TTL's as
high as 48 hours (such as Earthlink). So it wouldn't be perfect, but
it would prob
A potential scenario:
Two web servers with DNS, located in
different data centers on different networks. Each DNS server resolves the domain name only to its local IP.
Params in the zone files' SOAs are set
to very short expire times (minutes, perhaps). The registrar record lists
the two D
26 matches
Mail list logo
