The tweak that broke

Yes, I had a little accident ;-P

With the increase in technology came an increase in what I'd like to do with that technology, specifically be able to offer secure (https://) connections for security purposes on some of the sites I host. And although there have been moves to make virtual hosting of port 443 secure sites possible, they rely on changes to browsers which — as we all know — are a bit like a brick swimming against the prevailing current. Instead, therefore, just as with the ten year old specification of IPv6, we are required to use one IP (v4) address for each secure connection we wish to enable. So I obtained an increase in my IP allocation from a /29 (five effective addresses) to a /28 (with thirteen). Plans were made for a nice, smooth transition using proper scheduling of DNS and MX record changes when … I accidentally pulled the power on the modem-router and the change happened instantly. Without any of the advance setup.

Needless to say I was grateful for being the only person around at that moment as the world was filled with my choice 'language' about the unfortunate event, and I set to in trying to recover the systems. And, for the most part, I got the connections working again within the hour save for waiting for DNS servers around the world to play catch-up.

But e-mail wasn't arriving. My (Zimbra-based) mail server cold send messages out fine, but nothing seemed to be coming inbound. Eventually this was traced to a configuration error with my carrier this afternoon and they reset their end of things. Great? Well, no actually. At this end I lost all connectivity entirely. The outside world could see my router, yet I couldn't get beyond it. Three hours later the software-hardware-firmware-network interactions were finally resolved, I'm happy to say, so now everything should by A1 Bristol fashion.

That three hours without any connection was a bit scary though …

Back on-line

Did you miss me?

Over the last few days I've been upgrading the server kit I use to run my websites and API services on. The main server moved up from a dual 2.4GHz Xeon to dual 2.8Ghx Xeons, and from 2Gb of RAM tripled to 6Gb. I think that should cope with almost anything that gets thrown at it. (It is presently running around 50 websites and 3 million rows of MySQL data).

I also hived off my public NTP and DNS functions on to a separate box, and after reading some time back about how FreeBSD is a far better operating system for running the NTP daemon that is what the box running these services now is using. And I have to say the improvement in stability is amazing; something around a factor of 10⁴.

There are a few more tweaks to come, but glad to be back.