Mostly because I was thinking of implementing something for E2, and wanted a prototype, I've implemented a perfunctory mobile interface for modern mobile browsers (small screens, low-bandwidth for faster load times, and assuming HTML).

For the moment it's controlled by a per-client switch (in a cookie) so I can test it on desktop too:

[set_mobile]

Willow went down last night at about 1am due to the controller on one of the RAID drives crashing. The RAID subsystem quite sensibly removed the drive from the volume set, but I, having determined that it was probably transient, removed the drive from the set and attempted to re-add... without resetting the drive in the interim. After a reboot, everything seems fine and the array is currently rebuilding.

Needless to say, I'll be watching that drive like a hawk from now on.

We may be about to have a little bit of downtime. The problem with relying on a domestic electricity supply is that when you have an electricial round to fix a socket, the whole kit and kaboodle has to come down... sigh.

Or, I could just plug the modem in to the UPS as well...

Edit: Oh, it turns out it wasn't necessary afterall. Hooray.

Furtherly to the UPS being bought, installed and all that, this morning I took delivery of a 250GB disk drive which is going to live in Alder, and serve as an almost complete backup of most of what's on willow, and later on today will attempt to install it.

Potential issues abound. Firstly, alder is about 3 years old, I think, which makes me wonder if it'll actually recognise a disk larger than 160Gb. If it doesn't, I'll have to find some floppies to attempt a BIOS update. Assuming that will actually fix the problem, that is. If not... things get complex since there aren't enough IDE channels in Alder to support a 2x120Gb solution (swapping drives out of Willow into Alder and replacing those drives with the single 250), and if I were to attempt to rectify this by removing the extra controller card from Willow and installing in Alder, then willow wouldn't have enough channels (it'd require only one more than the mainboard has, whereas at the moment it uses two channels from the controller as well).

Assuming Alder is fine with the drive, I'll have to decide whether to opt for a fresh new Linux install on that drive (enabling me to swap it out of the system for safe keeping and stuff...) or to install it permanently and use an existing Linux install. Ideally I'd also like to mirror some of the server functionality too, to some degree, so that should the worst happen, there's a (horribly manual) failover path to get the house back online.

Mood: cautious

So. The journal mechanism here is a bit clunky. It's ecore'ish, in the sense that, like the traditional ecore 'weblog' mechanism, it requires that a document be created, and then linked to the journal it's associated with.

While here this allows any sort of node to be linked to a journal, the linking process is somewhat clunky. It allows nodes to 'fall between the cracks' if you create it and then forget to link it.

To try and get aronud this a bit, I've created the [journal update page], which will create an 'ljdoc' node and link it to a journal. It's sorta tied to the 'ljdoc' type, unfortunately. A more general solution would be nice, but... well, I'm not quite sure what the correct general solution would be yet.

For journals that you can post to, along with the usual 'edit' and 'rss' links at the top there should now be an 'update' link, which goes to the relevant page.

Mood: anxious

Music: Nouvelle Vague: Love Will Tear Us Apart

What with someone on edinburgers mentioning power outages last week, and now LiveJournal being somewhat screwed due to a power outage, I've taken what seems like a fairly logical step and ordered a UPS from Dabs.

It's a 320KJ unit, which ought to keep willow going for about half an hour after an outage. Enough time to cover most outages, I'd imagine, and shut down cleanly.

It seems like a fairly sane step, given the fact that they're building houses next door, and also that the weather is pretty rough.

Having said this, of course, it's entirely likely that there will now be a power "event" between now and whenever I actually manage to install the UPS, because fate has a keen sense of irony like that.

Prompted by a couple of random sysadmin tasks today, I installed Usermin. This lets folks with a system account manage, upload, download files, read mail, change passwords, and stuff like that.

Located at https://kahani.org:20000, y'all can use the new [Usermin functions] nodelet for quick links to the most important stuff, although Usermin's inbuilt sanity checking quite helpfully-but-annoyingly inserts an interstitial checkpoint when following the links.

Okay, new feature, kids. Journals now have the ability to automatically reassemble themselves as chronological aggregates of other journals. This is sorta like al :J friends list in effect, and is currently used for the new 'front page', which is an aggregate, in chronological order, of My journal, Lila's journal, Dermot's, and the 'news' journal, which is the old front page.

Haven't exposed the functionality much in edit pages or anything yet, and there's a distinct chance that if we have more than one such journal in the system it'll make the server explode and melt down. More later!

...and relax.

After spending most of the weekend faffing around, I've finally got a shiny new system up and running. Exactly the same hardware as Talula, bar the dead disk drive, but in my naming scheme, re-installing an OS on a machine is generally an appropriate time for a rename, since the OS defines the identity and character of the machine.

So, welcome to willow.kahani.org.

The name was Dot's suggestion, again. It's a name we've had floating around for a while, since it's what we almost named the cat. You can fill your mind with images of Warwick Davis, Allyson Hannigan, or trees, but it wasn't really named after any of these. It's abstract. (which is good, because I don't generally tend to like servers to have gender identities)

I seem to have subtly borked the postfix or Mail::Sender configuration though, so mail delivery from ecore just plain doesn't work. Apart from that, everything seems to be in working order. Huzzah!

So, I'd had this brilliant little idea a few weeks ago.

I had a bunch of packing foam from an iBook shipping box that my iBook went back to Apple in once. And it was pretty good at deadening sound. And I had this fileserver, see, that could use a bit of soundproofing.

So, yesterday, after upgrading the RAM in the machine, I cut some of this stuff to size to cover the side panel, closed it up, and was quite disappointed with the small amount by which the sound was muffled.

"Well," I said to myself, "I tried." And I left it at that.

I also failed to realise that the packing material was actually packed quite closely against the disk array. Until around 6pm today, when I noticed that things were subtly wrong. ldconfig would do nothing other than report bus errors. After some head-scratching I checked the system log and found lots of drive controller errors from the root drive.

Bollocks.

So, the root drive seems to be, for the time being at least, something of a write-off. Luckily nothing of it was being used other than 4Gb out of its 120Gb, but it was still the root partition. Some rigorous fsck'ing seems to have restored it to something almost usable, so I'll plug it back into the network tonight, but only as a peer on the network, not gateway. In the meantime I'll be building a new system onto a spare 10G disk as a replacement, to be dubbed 'willow'. I'll also take the opportunity to move to kernel 2.6.

So, in summary: downtime! Followed later by... more downtime! Hurrah.

Mood: annoyed