IPv6 down

It being a 'season'able few days where one isn't necessarily pushed to do work stuff, and not being at home for a few days, I hadn't checked that some of my services were running.

I just did :-(

Seems that my London gateway server's IPv4 connection is all happy and content but its IPv6 one is not. Doing the SSH thing suggests that I can't resolve it remotely as the external tunnel turns out to depend on the live state of the internal interface. Somewhat weird and unexpected behavior, but the drop in IPv6 connectivity appears to be the same time I turned off the internal routers (which I do if I'm going to be away a while).

Guess I'll never be turning them off again then!

b0rkken

Yes, I discovered earlier today that I had broken my install here. Mea Culpa.

I hadn't realised earlier as I hadn't tried to add anything here for a while as I've been so busy and, not too surprisingly I suppose, I hadn't gone seeking any of the other pages on the site.

Anyway, when I updated the security setups on each of my machines — as part of a general upgrade of everything to be IPv6-compliant — I'd locked-down some of the options available within Apache, believing that none of the sites I operate required them. Seems I was wrong, however!

I've taken a temporary measure to make things available again and will find a different solution later to sort the original issue.

Oh, and yes, every site is now available through IPv6 as well as IPv4. Yay!

Task done.

A while ago I started on what has turned out to be the long process of making everything I do available via IPv6. This evening I'm as far as I can go!

My initial problem was in getting a working IPv6 connection to my servers. I purchased a Cisco 1721 via eBay, and upgraded it with extra RAM and an interface card for ADSL. That, though, turned out to be the easy part of the process. Cisco routers like the 1721 use Cisco IOS as their operating system and though there are a number of sites online which have information about scripting the connections they aren't actually, well, 'complete'. In the end I was using guidance from New Zealand, Canada, and many other places to get some semblance of the right code to make it work.

Except it didn't.

Nothing would squeeze through my pipe, neither IPv4 nor IPv6. Eventually, one evening I decided on a 'do or die' approach (or rather 'stay awake until done no matter how long it takes') and managed to find some useful assistance from a tech support bod at one of my ISPs who tested from his end as I was testing from mine. In the end we had that sort of breakthrough like they had when digging the channel tunnel as suddenly the floodgates opened and traffic was passing this way and that.

But the next step was getting the various DNS services I make use of to have operating AAAA records so that my newly-IPv6'able websites could be reached. In two cases this required me asking the registrars to actually support IPv6 records (and I'm still waiting on one of them). Then, once I allocated all the addresses I needed I had to go through and update the host records for all the different sites and sub-sites. Which was a few more than I had thought it was. Over a hundred, in fact. But about 30 minutes ago that task too was complete and, as far as I can tell, they are all working just fine now, whether being accessed the 'old-fashioned' IPv4 way, or via the 'new-fangled' IPv6.

And most people will have absolutely no idea that anything has changed! Ah, the joys of being a tech geek.

Time

My public timeserver has been playing up a bit lately, but I've not been certain whether it is a machine hardware issue or a weather-related ADSL networking issue. As I've been pondering changing the machine which I use for the NTP service I've now done so, putting it on to a SunFire v100.

They'll be some stability issues for the next few hours as it stabilises, but I'm hoping it will be better after that.

Too cold for comfort

What with winter starting so early this year — and the snow falling and making the landscape all pretty like — bright orange lights have appeared (some flashing) on most of my servers. On two of them there are also error message codes which, going to the manufacturer's website to discern their meaning, tells me that the motherboard temperature is out of operating range. The site also tells me how to cool the machine down, for example by checking that all eight of the internal fans are operating properly.

Except my servers and the server room are all to clearly not too warm! Further investigation though shows that too cold would also result in the same error message and, after finding a thermometer and waiting for it to settle, I find that yes, indeed, my server room is at the bottom end of the 'ok' operating conditions at only 10° C. Poor things are shivering!

And now it has dropped a little further to only 5° C — eep!

I've tried insulating the cluster a bit but it is looking rapidly likely that I shall need to insulate for warmth the whole server room — and when we get back towards summer have to take the insulation out again to stop over-heating. A trip to B&Q seems to be on the cards.