Things have been fairly quiet on this blog recently. That has primarily been because things continue to work as expected, and there haven't been any significant changes behind the scenes.
However there has been some nice feedback, and a few new users have joined up pretty much soley as the result of the exposure generated by the recent Slashdot article:
That article was almost designed for small companies to advertise themselves, and so I did!
In all fairness I didn't go overboard, didn't make excessive claims, and there were many other companies being mentioned much more than we were. If you're looking for an alternative anti-spam provider there were many mentioned there, people such as:
- Antibody MX
- Barracuda
- MessageLabs
- Postini
Still I think I did a reasonably good job of promoting myself without being excessive - and it was nice to see that I wasn't even the first to mention the name!
Finally on to the real news, at the start of June we will have a secondary MX machine fully hooked up. I'll be making changes to the main server to smooth the transition over the coming days.
As part of the transition there may be small periods of downtime, but certainly no more than 5 minutes at a time - I just need to ensure that things will seemlessly fallover to the second machine, and that there will be no suprises when I advertise backup.mail-scanning.com.
For current users not following this blog I will announce the new IP/name via email to give everybody the chance to start using it.
Mail-Scanning Blog
Entries tagged "secondary".
As previously mentioned we're in the process of bringing a secondary MX online.
As part of this expansion the configuration of the primary MX (incoming.mail-scanning.com) has undergone many significant changes.
(These changes are mostly internal, moving away from a centralised database to one that may be replicated to an arbitrary number of secondaries. I'm happy to share details if there is any interest just mail me privately.)
The final step prior to the migration will occur tomorrow morning, and will involve around ten minutes downtime to the primary and the website.
Once that has completed the secondary will undergo testing for a couple of days to ensure it works appropriately, and then will be made available to all current and future users.
The final preparation of the secondary MX machine is now almost complete. To ensure that nobody was left waiting I delayed sending out this months invoices until I could be sure it would be ready.
If you're like to test your mail against the new MX machine please do, the address to use is:
- backup.mail-scanning.com
This machine is a completely independant from the primary MX, hosted in a different data-center on a different continent.
In the event that the primary MX machine fails the second should be able to continue working with a couple of caveats:
- Changes may not be made to the settings for a domain without manual intervention.
- The quarantine will no longer be updated in real-time.
Any changes you make to your domain will be reflected by the secondary within five minutes, and any mail which it rejects will make its way to the quarantine area within ten minutes. (There is a little lag to make sure there is not so much communication between master & secondary that either becomes unavailable for actually processing email).
Within the next few days the website will be updated to refer to both machines, and the support section of the site itself will be overhauled completely to make it more useful, and easier to use. Currently the backup MX has been tested as much as it can be, but there might be minor issues which only become obvious when the machine is used "for real".
I anticipate adding additional MX machines in the futures will take only a matter of hours; the migration to a completely "split" infrastructure required some significant architectural changes to the back-end, not to mention a lot of testing, both privately and with simulated traffic.
The secondary MX machine is now fully deployed and in use by over half our users; no last minute panic at all when the switchover was made, so I'm feeling pretty good about things right now!
The initial deployment of this additional machine was much slower than expected, because I needed to make significant changes to the way our service worked - It is pretty easy to setup a cluster these days, even a dynamic one using a shared/replicated database.
But the code that we run is a lot more involved than that. The ultimate goal of the secondary MX was to configure a box that would be able to handle mail processing in the event of the complete failure of the primary machine.
With that idea in place it became an interesting project to complete in its own right, and thanks to some brief moments of downtime I even demonstrated that this was possible.
Thankfully now that the second machine is in place adding a third, fourth, or even fifth machine will be trivial. Although I might have to come up with a naming policy for the new machines!
To celebrate the arrival of the secondary machine I've created a status page which will show the state of both boxes:
(As with this blog, and our support system, this status check is hosted on a completely independant third machine :)