Am Do, den 19.05.2005 schrieb Steffen Kluge um 11:05: > On Mon, 2005-05-16 at 22:00 +0200, Alexander Dalloz wrote: > > > By using a server with a lower MX priority. > > > > But how is that transparent to the clients? That was a criteria for > > Cosme. And I think he means clients trying to send mail. > > If the lesser preferred MX is working, the message won't be deferred and > no warning will be sent to the original sender. > > > If > > mx1.domain.tld is off which is used by the clients for SMTP, how does it > > automatically switch to mx2.domain.tld without user intervention? > > "It" meaning the SMTP client? It just does, at least all SMTP clients I > had the pleasure of managing so far do. E.g. from the postfix smtp(8) > man page: > > The SMTP client looks up a list of mail exchanger addresses for > the destination host, sorts the list by preference, and connects > to each listed address until it finds a server that responds. > > Only if it doesn't find one it can connect to, or if an error occurs > after it did connect, the message will be deferred and retried, with > notifications starting to go out after 4 hours, etc. > > > But as said before in this thread: properly configured MTAs will > > retry to deliver up to 5 days at least. > > But not everybody wants to wait several days for their emails. Over > here, people start jumping up and down if there's more than 30 minutes > delay... > > In my experience, the SMTP/DNS combo is a pretty good HA design all in > itself. Just use multiple identically configured mail servers, and give > those you don't "normally" want used a lower MX preference. The rest > comes automatically, completely transparent to email senders and > recipients. > > Cheers > Steffen. Thanks Steffen for commenting. My point wasn't speaking about foreign MTAs to deliver to any of the responsible MX hosts. I think I pretty good know how the scenario works :) "I think it is not a good idea to allow my client see a warning about my server unavailability." I interpreted this sentence as Cosme speaks about his customers or users in general. You can't avoid that problem with a secondary MX for the users when they try to SMTP send with their client through your MTA, when it is down. And who is not able to read and understand the common warning message after 4 hours of delivery failures (warning, not error), the person should not use email. But a real argument: if the primary MX is that unreliable that it is that often down or however unreachable - short time for hours or even days - it is of course much better to change it's connectivity than to add another host and raising up the costs for administration. The argument by John Summerfield about the mail access server being offline when the MTA is off is as well a good one. Getting a robust and transparent HA mail server system (MTA and mail access) is much more than to add additional MXs. Though people may have the impression email is a real-time service - it is not. There are enough big mail service players with very long delivery retry time delays. Alexander -- Alexander Dalloz | Enger, Germany | GPG http://pgp.mit.edu 0xB366A773 legal statement: http://www.uni-x.org/legal.html Fedora Core 2 GNU/Linux on Athlon with kernel 2.6.11-1.14_FC2smp Serendipity 17:08:37 up 5 days, 16:40, load average: 0.01, 0.05, 0.04
Attachment:
signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil