varunnangia.com

I posted this call for help on SmallNetBuilder.com, since that’s one of the smartest, most knowledgeable communities on the Internet for finding out what to do with small, home networks. But I figure I ought to cross-post it here too, in the hopes that someone else might end up here instead. Your help is appreciated!

I came over here looking to just mention my experience with FreeNAS the last few weeks, and lo – Tim’s asking about RAID recovery. Slightly freaky.

I had my FreeNAS box go south on me over the last couple of days and I thought I’d just share two things I’ve learnt from the experience. A little background first: for over a year, I’ve been running FreeNAS 0.6x on a home-built Celeron 430-based machine, four SATA drives for storage via software RAID, a throwaway IDE drive for the OS. The basic setup was two RAID1 arrays joined together as a JBOD, a decision I have come to regret, but thus it was, thus it is. So what have I learnt?

1. RAID recovery does not work. There is inevitably something that will go wrong. In my case, two things went horribly wrong. First, even though I could read FreeNAS’ native UFS formatted file system, I couldn’t recover any of the disks because FreeNAS partitions using the GPT partition structure, and NOT the more common MBR partition structure. Second, because of my decision to JBOD the arrays, the file system was partially on the first array, and partially on the second. No recovery software I found could handle this setup.

2. RAID is not a backup, so backup, backup, backup, or don’t give a damn. Because of my inherent paranoia, I’ve been copying stuff from other computers on to the NAS, in the hopes that everything doesn’t go south all at the same time. The net result of the NAS crash is that we’ve lost: my current resume (and I’m pretty sure I have it in my email somewhere), the current network map (not that difficult to recreate, as I have older ones around), and two bits of software (which I can re-download and re-request keys for). I’m sure we’ll discover a few other things that have gone missing over a few weeks, but it’s not bad. So data loss is not the issue here – but time and convenience is. It was very simple for me to go to my software share and reinstall some piece of software, or to go to the music share and stream some music. That’s all on hold. Moreover, I’m currently missing a safety net, in case something else should go sour.

I’m going to give some thought as to what to do next – and your comments are welcome and appreciated. One option that I am going to rule out before someone suggests it is reinstall FreeNAS. This is the third time in as many years that I’ve had something go abruptly wrong, with no explanation, even though it’s apparently a known issue in the current nightly releases, and must exist earlier as I was running an earlier build. My time is not worth another FreeNAS install, though I’ve loved using it when things are going well. The fact is, there is no reason an orderly shut-down should result in chaos when brought back up again.

Options that I’ve come up with so far:
1. OpenFiler. I’m told this is a painful OS to administer, but a pleasure to use. I’m not sure I want to necessarily learn yet another OS to just have my files available everywhere. On the other hand, my previous experience indicates that it supports a significantly larger set of features than FreeNAS, including one thing that FreeNAS fails at – the ability to use (slightly) different size disks.

2. Windows Home Server. Testing here and elsewhere indicates that WHS is faster than its alternatives, and very user-friendly. However, the lack of a web-based administration front-end means that I’m stuck to using a PC running Windows to administer it, which is not an option. Microsoft also rubs me the wrong way philosophically, though I’m willing to put aside monetary and philosophical concerns if the damn thing just works and allows me to upgrade as larger hard drives become available.

3. Buy a BYOD NAS like the QNAP or Synology devices. Yes, that’s nice. It’s also several hundred dollars, and locking into a storage vendor’s options. What I mean by this last thing: codec support missing for video files, or some proprietary file system, or, worst of all, a feature that’s locked out because you’ve not paid for some annual subscription. Yes, there are relatively standard x86 computers inside some of them and no, I’m still not convinced. Unless someone has had a successful experience recovering a device, the trade-off is not worth it.

4. Forget it. No network-based storage. Expand local storage with a local RAID (or RAID-like) array, and use my Mac OS X based machine as a server to share out the files. I’m most tempted by this option – where this means buying a Drobo. I’ve heard both good things and bad things about it – no one I know has ever had an issue with their Drobo, but there are enough horror stories on the internet about Drobos going south that I don’t know whether I’m living in a statistical fluke.

5. Something else. This is where you all come in. Surely this can’t be the entirety of the options. You’ll notice that all of the options are those that keep the device (and data!) under my control; that’s a (strong) preference, but not an absolute requirement – provided strong encryption and incremental backup are both possible.

Thanks for listening. And remember to do your backups.

One of the things that I’ve been debating about is what to do with older, perfectly serviceable, hardware. It’s a problem that’s become worse in recent days with the development of the Intel Atom, dual and quad core chips, but it’s not recent – this is a problem that’s followed me for a while, but of late it’s become a little ridiculous.

It started with a friend in college dropping off a 1.0GHz Pentium 3 box, saying that he couldn’t get it working, and had bought a new computer. As it was the middle of the exams in the fall quarter, I put it aside, and figured I’d get to it a little later, maybe after winter break. Opening the case after the break revealed that the machine was fine, but that dust had clogged up the fan, preventing it from spinning and causing the BIOS to refuse to turn on the computer. 15 minutes and a through brushing later, it was up and running and I called my friend – only to discover he’d graduated early and that he no longer wanted the computer back. “Cool. Free hardware!” I thought. I decided to put Windows Server 2003 on it and learn the basics ahead of a major transition to Server 2003 from NT 4 Server and Server 2000 machines at a city school I volunteered at. A year later, I installed an early version of Ubuntu (I think Hoary Hedgehog, but it may have even been Warty), and used it to become familiar with Linux. A year after that, I sold it at graduation for a $100, a princely sum for a machine that I acquired for free and was the better part of a decade old by then.

Two months after graduation, a friend came visiting me from New York, bearing gifts from another friend – an apparently dead AMD Athlon64, motherboard and an ATI Radeon 9800. Some amount of twiddling later I figured out that it was a bad capacitor on the motherboard, but I had neither the tools nor the soldering skills to replace the component. Instead, two years after I got it, I passed it on to a colleague at work who was studying for his CCNA, and told him how to repair it.

Of course, in the interim, I had acquired quite a collection of hardware from various sources; mixing and matching produced my current home server from spare parts, and a half-done photo-frame PC. Other parts have made my home network, allowed me to fix nearly a half-dozen laptops for friends and relatives and so on. But I still have a ridiculous number of parts from all sources, ranging from the useful-by-itself (a dual-core Athlon64 x2 HP Slimline with a bad 6150 that will become my and my flatmate’s DVR after the digital transition is done and we rid ourselves of cable) to the what-do-I-do-with-this (a Mini-ITX board with a soldered down ULV 600MHz Celeron that served as a firewall). I’ve tried selling a number of these things, but in this economy, no one’s buying, and those things that I’m willing to give away, people don’t see enough value in coming to pick up.

Most recently, two days ago, I resurrected my flatmate’s “obsolete” HP machine and made it a dedicated encoder machine – it sits and converts media from useless formats to useful ones. Sure, it’s slow, but I can’t help but feel bad about tossing a 5 year old, 2.0GHz AMD Athlon64, 400GB of hard drive space and a gigabyte of RAM. I’m still not sure what to do with the little Mini-ITX motherboard; any suggestions? What about a 1.6GHz Pentium 4M laptop? What about a 1.42GHz G4 Mac Mini? So far, I’ve thought of an Asterisk server (don’t really need it), a development box (for what?), and a bedside computer (though I have a netbook) respectively for each of those. Or maybe I should send the Mac Mini to Nevada, and give the laptop to my parents?

One of the pretty nifty new features of Ubuntu 9.04 is its global notification system. Just like Growl for Mac OS (and clone Snarl for Windows), this basically creates a single, unified, neat little notification system that shows up in one spot on the desktop. Not only does this cut down on clutter significantly (just take a look at the mess that is the Windows task tray notifications, if you don’t believe me), but it also makes the whole OS less intrusive, yet highly communicative.

It got me thinking, though, about a universal notification system – as in, for every computing thing in the world. That includes things like “Machine with MAC ID 00:11:22:33:44:55 (‘Varun-Nangias-Mac’) connected to the router” to “Updates available”. In fact, most of the messages (“New SMS on Blackberry”, “Printer on fire”, “New Yahoo! IM from Kats Gupta“, etc.) all fit in 140 characters. Sure you could have longer messages but for the most part, 140 characters is more than enough to communicate the jist of the message and to provide a no-nonsense summary of the event; if you really need more information, you should be able to take further action on by looking up a syslog, or the requisite application. And putting this information in an RSS feed or a Twitter-like stream on the Internet means that it’s easily accessible from almost any device with a web browser.

Given that Laconi.ca already exists and can be deployed to your own server, basically what is missing is software that links clients to the Laconica backend. Ubuntu and Mac OS already have most of that support – the notification systems can forward events that they receive. Windows support is sort of there in the form of Snarl (though very, very few applications use it). A J2ME application is needed for the vast majority of phones out there; an Android phone can have this sort of background daemon running all the time, anyway. Apple would likely need to build some sort of support into its push notification system in order for iPhones to mimic this functionality.

Put all of this together, and you’ve suddenly got a way to aggregate your digital life into a single, very easy, stream of information. Add the ability to respond to certain events via replies (ex.: “Updates available. Reply ‘install’ to install available updates”) and the cost of your remote management system has dropped to pretty much zero. How cool would that be?

This time last week, DC was under a snow watch, with about 20cm of snow expected, temperatures around -5 (and dropping), and the airports closed (which was a real problem). Today, it was 25, and its not quite the equinox yet. All signs point to a pretty brutal summer ahead and I’ve started developing coping strategies – both for myself and my cat, and my machines. For the living creatures, this is pretty simple: cold water; but the computers present a problem.

Last week, I was using the machines to provide some heat by leaving them on; today all the computers (save for one) are silently suspended to disk. The problem is the one that’s still on – my home server. While it caught the tail end of summer last year, since then, more drives have been added and the amount of work its doing has increased. Last year, each disk was reporting a temperature of about 46-47 degrees; this year, the added drives and load is already resulting in disk temperatures of 50+ degrees. This is despite adding another fan to the case, and improving the ventilation where the system is located.

I’m tempeted to do to it what I presecribe for humans – cold water – but it both seems silly and unnecessarily risky (not to mention, expensive) to take a chance with my home server getting drenched because of a water leak. So, I ask: any suggestions? Something relatively inexpensive I can do to improve the temperature of the case? Some things I’m considering: replacing the steel case with an aluminum one; replacing the 35W Celeron with a 12W Atom, replacing the two sets of 500GB drives with one set of 1TB, disconnecting any drives not actively being used, setting up the system to sleep and WoL if it hears a device calling for it… But I would like to take the server down once only to implement all the changes… so any suggestions?

So, it’s that time of life again – time to buy a new computer. Since the last time I bought a computer, the options have become infinitely more varied and difficult, so if you’re interested in helping me choose a computer, click through. It’s not a long post, but it does require some thought.

Read the rest of this entry »