Sort of stranded at my parents'

May 15, 2006 / 0 comments

I went back to Northern Ireland for the first week of May. Vicky and Branwell were getting married on the 5th, so I figured I'd go home the weekend before (with the bank holiday), work a 3 day week, then come back on the Monday. Plenty of time to see my folks, some friends, and relax.

I did end up feeling a little stranded at points. My parents live in Newry, and I don't know anyone there any more. Public transport in NI is dire and I didn't have access to a car (even if I'd been insured for Mum's she's far too busy for me to be able to borrow it for any length of time). Kathy was in Banbridge (15 miles away) and there were people in Belfast (40 miles away) but it wasn't like I could just pop out and see them.

However, what I did realise is that there's a social system in place back "at home" that I could easily slip back into. I spent the Bank Holiday afternoon pottering round Belfast with Jake, Paddy and Peter, ending up in The Kitchen. It was, as it always is, good to see them. Later I went out with Kathy, Mal and Miranda for dinner. At Vicky and Branwell's wedding I saw a bunch of people I probably haven't hung out with in 5+ years, all of who were chatty and friendly; the wedding was a good day/night. I didn't have time to see ct, though it would have been good to.

I'm sure I could slot back into a similar setup in Oxford, but I found it surprising and comforting to realise I still have that in Northern Ireland.

E3 booting without help.

May 14, 2006 / 0 comments

I've been quiet for a couple of weeks now; I went home to my parents for just over a week at the start of May (more on that later), then last week was spent catching up with various bits after being away, and then the past few days I've had a cold (just in time for the weekend, yay!). Meh.

Anyway, as such I haven't been fiddling a lot with the E3; still stuck at the point where I could get it to load my second stage loader from NAND, which would then load u-boot, which would then load the kernel, but not getting anything from the kernel after decompression. Whereas if I loaded the second stage loader over the serial port it all worked fine.

Figuring that Amstrad's PBL must be doing something different depending on which method you use (NAND/serial) I had a look at the config registers to see if anything was different. Nothing obvious. I enable the low level kernel debug output and find it is booting ok, but that it freezes as soon as it tries mutex_lock in clk_register. Odd. So I disable CONFIG_DEBUG_MUTEXES and all of a sudden it works. My E3 boots into Linux unattended. Woooo. I think.

Xinit considered harmful.

May 4, 2006 / 0 comments

Last year Black Cat decided they wanted to be more organised about offering dedicated servers. We already had a rackful but they were all in use. So we looked around for a suitable supplier. Originally Dell were the front runner; although they have a bad reputation we already had a bunch of machines from them and they'd been reasonable reliable, did serial BIOS properly and were competitively priced.

Dom pointed out that having hotswap drive bays would be a really good plan; we'd already specced the machines with dual drives to enable RAID1, but if they were hotswap then we could easily move drives into a spare chassis in the event something went wrong, meaning the customer could be back up and running again quickly while we investigated the old machine. Good plan. This started to push Dell into ridiculous pricing, so we had a look around. We had some machines from Xinit, which again we hadn't had issues with. They were able to give us a price that wasn't that different to the non hotswap Dells, so we decided to go ahead with them.

Bad move.

On September 6th I informed them we wanted to go ahead. They'd quoted us a 10-14 working day lead time. As we didn't have a credit account already set up with them and it was likely to take a while to sort out we said we'd pay up front in the interest of getting things moving as quickly as possible.

By September 22nd we still hadn't heard anything from them, like even confirmation that payment had been received and an ETA for delivery. So I contacted them, received confirmation they'd received full payment by September 13th and was told we should see systems by September 30th or October 3rd at worst.

On September 26th they emailed to say there was a problem with their shipment and we'd be receiving a part shipment with the rest to follow.

4 servers were delivered on October 12th. It turned out none of them had been configured as requested (serial console, 9600 8n1, all set to auto power on after power failure), so I ended up having to plug a monitor and keyboard into each machine before connecting it up to the serial console server and power cycler.

I chased Xinit again on 24th October. They'd claimed a 5th server was to be delivered before that and still hadn't told me anything about the remaining 15. I was not impressed. A new contact got back in touch and made all manner of apologies and promises about keeping in better contact and keeping me up to date with what was happening. Suffice to say this didn't happen, and calls to the main Xinit number were of no use as no one could help, and the mobile number I'd been given for this contact was unanswered more often than not.

On 1st November we received another 11 machines. These were at least configured as requested, but during our testing we discovered they had been misassembled. 5 machines had CPU heatsinks put on incorrectly; visually misaligned and such that any load caused a machine crash. Very unimpressive from a company who supposedly does full burn in tests on machines before shipping them (and indeed used this as an excuse for not shipping us machines sooner on occasion).

Further chasing didn't have much effect, but we eventually received the remaining servers on December 5th. Nearly 3 months after we initially placed the order.

While Xinit have happily provided many platitudes they have failed to address the numerous failings in their organisation. They are appalling at responding to phone calls or email. They continuously fail to make good on any of the promises they make. They can't provide realistic delivery dates and don't have the decency to get in contact and let you know when they're going to miss them. They have proved unable to assembly machines to any level of competence; most of what they do is buy in chassis and fit parts to them - this is not complicated and yet they managed to get it wrong on several of the machines we ordered. We were promised a machine or two in compensation, but of course these never arrived. In addition, they're so badly organised that in the middle of March we received a spam from one of their sales people who was completely unaware we'd been customers of them!

I cannot stress how much you should never consider using this company. I've spoken to others who've bought machines from them or had dealings with them and I know I'm not alone in this. Really. Don't do it. You'll regret it. I suspect it would have taken less of my time and energy to buy all the parts and assemble the machines myself. I wish I'd done so instead.

This does leave me lacking a hardware supplier however. We're in the position of needing more machines, but given recent power pricing increases Xeon LVs (the new Sossaman cores) look like the way forward. Except no one appears to have any stock. And even if they do, how do I know I won't get my fingers burned again?

You know you're old when...

May 3, 2006 / 0 comments

You sit next to a teacher at your brother's AS Level drama play, don't realise she's a teacher until someone tells you and then think it's highly likely she's younger than you or at least no older.

Anyone out there with a VT8235 datasheet?

Apr 27, 2006 / 0 comments

Argh. A long time ago I got LinuxBIOS up and running on a Via EPIA-M mobo - someone else had done the initial work, but it had bitrotted. As part of that I got auto power on working, such that if you removed power from the box it would come on when you reapplied it, even if it had been turned off before. However I never checked this into LinuxBIOS as it's a bit intrusive - not everyone wants boxen auto powering on. What I now find is that not only have I lost the code to do this, but the copy of the vt8235 (southbridge in use on the EPIA-M) datasheet doesn't have the details of which register I need to twiddle. I'm sure it used to, but that might have been an older copy which of course I no longer have.

So, does anyone out there have a copy of the vt8235 datasheet and can tell me what magic value I need to poke to what register(s)? Please? I've tried creatively Googling, I've tried running Via's factory BIOS and seeing if the PCI config registers changes when I flip the option in that, but all to no avail. Someone save me. I don't want to have to disassemble the BIOS trying to find this one tweak. :(

subscribe via RSS