Parte the Firste - 4am, 26th May
This site isn't responding, either to myself or to people in the US I've given the link for a page here to. The mail server is also unreachable. From a web browser or email client, anyway... there are a couple of quick tests to determine whether the thing is down completely: ping and traceroute. (Not all that interesting, I'll skip reproducing output.) Both work fine, whilst with a web browser the result is one of Firefox's friendly XUL error pages: Net Reset Error and "document contains no data." http://188.8.131.52/ and http://libra.plugsocket.com/ (aliases for the server the site is on) similarly result in timeouts, though ping and traceroute likewise work fine.
The last time I noticed was May 9th, at about 2am. Fired off a quick mail through the PlugSocket contact form and got a reply to my TFA address with usual quick service: "It's currently performing a disk check and should be back up within 30 minutes. Mail sent whilst the server is offline may be delayed slightly. Apologies for the inconvenience."
Fair enough. I was simply concerned that mail to and from the server could have been lost, as I was busy working things out with an American who'd bought a stack of magazines from me on eBay without taking into account international postage for 17.5Kg of paper. The test mail I sent from a throwaway Yahoo account during the unreachable period confirmed that that side of things was queuing correctly by getting through by the time I woke up.
I'm just wondering if the site being unavailable to visitors at this sort of time is a routine occurrence. That would be far from acceptable; few people in the UK will be paying a visit, but 2am here == early evening in the US or Canada. The web is global, and PlugSocket advertise 99.9% uptime. Of course, if we were splitting hairs, the server is up—it just isn't responding in a manner that allows documents to be served.
Some other quick detective work before I hit the sack... a blog elsewhere that links to my PlugSocket recommendation in October is working fine. That site's at 184.108.40.206, which reverses to eagle.plugsocket.com (other PlugSocket subdomains I know of include maple, flora, inkle, ns.flora, ns1, ns2 and cobra, all of which are also responsive.)
Parte the Seconde - 6pm, 27th May
Got a mail back this morning stating that the server is one leased from another company that they're experiencing unusual ongoing problems with. A transfer is possible if I want.
First of all, I appreciate the honesty. The last move was due to a server with limited configurability (no option to use .htaccess, mod_rewrite, etc) so I've fired off a mail to ask if a move would cause problems there. I don't do anything particularly complicated with .htaccess, just php error pages, domain redirection, remote image link blocking, etc. but these are reasonably crucial things to my mind.
Re-uploading everything again wouldn't be a problem as I currently have a broadband connection to play with. What might cause a bit of hassle is if the MySQL database used by the commenting systems here (and for a passworded forum most of you don't know about, which is a convening area for the TEFL team going to Poland with APASS in July) changes again... configuration for the forum may need to be updated, database names in my own scripts changed, etc. Parts of the site should be closed, or it risks losing comments or posts people make on the old server whilst DNS changes roll over.
(DNS? It's like a phone book for the internet. Nameservers around the world keep lists of domain names so that, for instance, when someone types www.tfarchive.com they get sent to the server at 220.127.116.11, and when DNS changes the nameservers have to update their lists. Someone in the US may not get their list updated as quickly as someone in the UK, so that when that person types in a URL they see the old server rather than the new one. The internet was designed for routing around damage rather than for speed: DNS propagation of nameserver switches can take days, though simpler changes made on the same server are more rapid.)
Parte the Thirde - 1am, 28th May
If you're reading this, you're seeing the moved site (seemingly at cycle.plugsocket.com at 18.104.22.168)—the blueyonder servers at Telewest have the new DNS, enabling me to witness this. Yahoo also have it, probably because they route their email service through whichever system is geographically closest to a person logging in. TFArchive doesn't (I think it's physically located somewhere in Texas) and JANET (the UK Joint Academic Network) doesn't, as mail from those addresses has just turned up over at the libra server. I'll stick the webmail for that in an auto-refreshing browser tab tomorrow.
All settings and everything got mirrored; apparently the configuration is the same on the owned server, which facilitated an effortless transfer. Thumbs up to PlugSocket, who are in a rather tricky position. They can't proactively transfer all accounts away from libra, because that would cause the problems with mail and interim site changes described above. They could get in touch with anyone else with an account on libra, if the problems haven't been resolved or there isn't a short timeframe for them being fully resolved, and schedule 'static' time for those sites. They can transfer post-haste anyone who reports problems, which has been the case here. All because some gits leased them a dodgy server. I really wouldn't like to run a hosting business...