How to reduce reliance on a single web host for email delivery?
August 6, 2004 3:31 AM   Subscribe

Web hosting woes: My host has had massive email issues for the last few days. Would it make sense to move DNS to somebody like Zonedit so that I can redirect email to other domains temporarily when situations like this arise? Any other suggestions?
posted by SteveInMaine to Computers & Internet (5 answers total)
Wouldn't the biggest issue be how long it takes the changes in MX records to propagate? I mean, wouldn't it be several hours, as a general rule, before all mail got re-directed to the new host, and then several hours to switch back afterwards?

Put it this way--if your host goes out for that long, often enough that it's worth doing this kind of workaround, then you should really be looking for a new host, I think. It's their job to minimize how outages affect you, not yours.
posted by LairBob at 4:51 AM on August 6, 2004

Assuming you have email service at other domains which could handle such a redirection w/o further configuration changes (e.g. if you have your homegrown server which only accepts mail for '', and this webhost is '', then telling DNS for siteB to go to siteA will cause your email to simply die on you!)...then yes, it's a great idea.

I use ZoneEdit for my domains and they update _fast_. I have a subdomain set up so I can SSH into my home computer (behind DSL) without having to remember the IP address. Whenever I have to reset my firewall (various reasons, mostly involving the DSL router being accidentally powered off), my IP address changes.

When that happens, I go to ZoneEdit's site and make the appropriate change; the propogation usually only takes an hour or two, sometimes only a few minutes! So I <3 ZoneEdit :)
posted by cyrusdogstar at 11:03 AM on August 6, 2004

Best answer: the propogation usually only takes an hour or two, sometimes only a few minutes!

That's the propagation to your ISP, not worldwide propagation. Update intervals vary, so while you're pulling mail from Server B there could easily still be mail dead-ending at Server A for days after it's "propagated" in your eyes. (more info)

In Sept, Verisign is planning to start doing continuous refreshes of registration data instead of just twice a day, which should help speed up that process. But apparently propagation of A and NS record changes will not have the same fast refresh as registration data.

Have you thought about switching hosting? Server problems happen to every hosting co, but "massive mail problems" shouldn't be stretching out for days. If that's happened more than once, I'd start shopping around.
posted by nakedcodemonkey at 11:56 AM on August 6, 2004

Response by poster: Well, here's story: There were intermittant problems with email Tuesday and Wednesday. They attributed the outages to large mailbox sizes. They applied an upgrade to the server (only applying the upgrade to the most problematic server for one of our 4 domains) Wednesday night, and say that the upgrade itself took 20 minutes, but then the software began a message conversion process which pointed out a problem with their disk RAID. Not wanting to shut down the mail server completely and sacrificing the message queue, they are now in the process of moving messages over, account by account, to a new drive array, and now say they expect to have everything back up to speed by 10 p.m. tonight.

Right now my main issue with changing hosts revolves around a bunch of .NET apps that we have running on one of their web servers. Although there should be no problems migrating to a new windows host, the transition would still take a couple of weeks to make sure nothing's broken.

I should point out that this host (Crystaltech) has had a rep as one of the top windows hosts and we've been with them for going on 4 years, so I'm trying to cut them a little slack. At the same time I'd like to have a fallback, so if this happens again I'd be in a position to continue getting emails and be able to quickly move to a new host if necessary. With an outage like this, I'm likely to move off their servers if this happens again. They've run out of chances as far as I'm concerned.

Right now another strike against them is they didn't anticipate the number of messages needing translation. They claim this server has 25k accounts and 2,000,000 messages. They're also pointing to one of the customers on this server who decided to send out 80,000 messages while the conversion was taking place. Though this is a boneheaded move on the customer's part, they were well within their rights as long as the action fit within the TOS.
posted by SteveInMaine at 12:44 PM on August 6, 2004

A bit late, but I'd like to add my love for zoneedit to the record here.
posted by John Kenneth Fisher at 7:59 PM on August 6, 2004

« Older Preventing Colon Cancer   |   MIDI. Through USB. In Linux. Newer »
This thread is closed to new comments.