Web Hosting Talk : Web Hosting Main Forums : Hosting Security and Technology : DNS MX Question : Multiple choice Question

[nipl]

I was wondering what would happen to incoming mail in this scenario...

webhost.com MX 10 supermail.com
webhost.com MX 20 backupmail.com


Choice A) Mail goes to supermail.com ONLY
If supermail.com is down, only then it goes to backupmail.com

Choice B) Mail goes to supermail.com AND backupmail.com


I believe ChoiceA is correct. But I need to implement ChoiceB. I was thinking of keeping both on same MX-Priority but not sure it will work perfectly.

Input from DNS experts needed on this !

[John[H4Y]]

Choice A

[matt2kjones]

yea choice a

if you set them to the same priority i think it will radomly go to a mail server

or maybe it will go to mail server one

it wont go to both

[nipl]

Yea, I think it will be random too...sort of like Round-Robin DNS.

Hmm....can I convert this to a subjective question now

Has anyone got any any ideas on how todo both mailserver simultaneously ?!

When my client put forth the requirement, I thought it was a pretty genuine request too...he didn't want any email missed ever. The 2nd backup mailserver won't be hosted with me ..but still I've got to advise on the solution for this.

Also, I think its a pretty cool solution to offer for redundancy of critical email...as long as network stays up, it can be offered on 2 separate mailservers by any hosting co. and important clients don't lose important business emails in case of a software or hardware glitch !

[nycmaster]

one question.

my understanding is that if you have:

webhost.com MX 10 supermail.com
webhost.com MX 20 backupmail.com

and supermail.com goes down. The email sent to you will que up at the senders mail server and will not automatically bounce to the second backupmail.com server.

Only after it fails ( anywhere from 1 - 4 days usually ), will it then look for a 2nd mx record to try.

or am I wrong ?

[nipl]

Yup, you're right...it stays at sender's mailserver. But when business email is really important and deadlines are at stake, the user won't wait for 4 days...hell not even for 4 hours !

And such business users will pay for whatever it takes to keep email running 100%...won't even settle for 99% and any kind of delays either.

With simultaneous delivery to backup mailserver, the user simply retrieves from there instead of sitting around biting their nails.

Come to think of it...when some co. has 20 employees working in their office, and the mailserver goes down...they lose productivity for the entire 20 employees. That means lost money for that Co.....and everyone knows salaries cost much much more than web hosting.

[nycmaster]

With simultaneous delivery to backup mailserver, the user simply retrieves from there instead of sitting around biting their nails.

that would be great. I've looked at doing this but never found the right answer.

ok - so how do you achieve this ?

if you set the priority to the same value of each server, won't it act like a round-roubin, where 50% of mail goes to #1, and the other 50% to #2 ?

So if #1 is down, only the other 50% will arrive at server #2 ?

no ?

[nipl]

Yes, that's what will happen....which I don't want.

We're in the same boat...I don't know the answer either...that's why I put this question to the WHT community because I'm stuck on this one. I spent a lot of time researching DNS resources...everything says MX Priorities work in sequence, not in parallel.

I would really jump with joy if someone says "its can be done"

[RofyHost]

Hello everyone,

Nope nipl, I think nycmaster is wrong cause if you have the following configuration:

webhost.com MX 10 supermail.com
webhost.com MX 20 backupmail.com

and supermail is not accepting connections (down) remote MTA should fall back to the secondary MX thus backupmail

no delays and no waiting for 4 days or hours

Greetings,

[John[H4Y]]

To do something CLOSE to simultaneously, you could use something like rsync to sync the mail spool files between the two machines. This obviously wouldn't be instant unless you ran rsnyc over and over. You could also sync them by remotely mounting an NFS filesystem from one to the other or something like that.
What would the point of having mail go to two places at once be anyway? As long as the backup mail server works when the primary is down, you never miss any mail.

[nycmaster]

John[H4Y] --

I think the idea is not not miss any email BUT also not to have a delay on email. I actually think that unless your outage is longer than 2-3 days ( very unlikely ), you won't actually miss any email with just 1 server.

what we are asking is how to have a situation where it's business-as-usual while one mail server is down.

rofyhost --

I'll do some tests, but I'm pretty sure the sender's mail server will keep trying the primary mx for a few days ...

[sprintserve]

An alternative is to use your settings, but instead of rsyncing, you do a mapping. i.e. all servers that is sent to mail.com are forwarded to their corresponding emails on backupmail.com. That may work, but can be a lot of work if you have lots of users. This should be less demanding on the server load as well.

[RofyHost]

hi again

well, there is also a good way to achieve what you want, I know MS Exchange supports Clustering, check if your mail server does! I think Clustering is the best solution.

Greetings,

[nycmaster]

one more data point.

I'm testing zoneedit.com's mail forwarding feature.

You can have a single or all email addresses of a specific domain forward to another email address.

So I tested taking down my primary mail server, and at the same time turning on the forwarding feature.

Within about 30 minutes, emails I was sending to the domain with the down mail server, began to forward to the new address.

so in case of a long outage, this might be another way to keep email flowing.

my next test is to swap MX records and place a secondary MX record as the primary, and see how long that takes ...

[xyzulu]

Why not make the mx values of both the same. Then mail should go to either server. As far as I know, that should work. But i stand to be corrected.