**** BEGIN LOGGING AT Sun Jun 24 02:59:58 2012 Jun 24 03:46:32 ogra_: Do you have a non-ES Panda that I could get you to install a natty kernel on for me to experiment with? Jun 24 03:46:59 janimo: Or you? Jun 24 03:47:23 GrueMaster: Or you, if you still have a bunch of hardware collecting dust. Jun 24 03:52:22 infinity: I have one Jun 24 03:54:34 stgraber: \o/ Jun 24 03:54:45 stgraber: Could you install a fresh natty headless on it and give me root? Jun 24 03:54:55 stgraber: And maybe attach a USB disk, so I don't go insane? Jun 24 03:55:59 stgraber: One way or another, I'm going to fix this &^#$% mono/armel thing. But, after my last round of fixes, I can no longer break it on my PandaES. :/ Jun 24 03:56:15 stgraber: (So, I'm blaming the natty kernel, until I can prove otherwise) Jun 24 03:56:55 infinity: installing natty should be easy, finding a usb disk will be trickier ;) Jun 24 04:00:20 the only usb storage I have around is my n900, not sure it's going to be a lot faster than the sdcard :) Jun 24 04:01:48 stgraber: If the SD is big enough to build mono on, that'll work, I suppose. Jun 24 04:02:01 stgraber: Unless you have an NFS server you could cut me a slice of. Jun 24 04:02:24 sure, nfs is easy Jun 24 04:02:50 Anything's better than SD. ;) Jun 24 04:03:29 Some sort of tiny-magnets-over-IP would do. Jun 24 04:04:00 * infinity knew he shouldn't have given Andy his old Panda. Jun 24 04:04:49 /dev/mapper/external-adam 3.7T 56G 3.5T 2% /data/external/adam Jun 24 04:04:55 ^ should be enough for mono Jun 24 04:05:01 I should hope so. Jun 24 04:05:17 infinity: got IPv6 over there? Jun 24 04:05:25 No, I live in the past. Jun 24 04:05:28 I should fix that. Jun 24 04:05:35 But not this evening. Jun 24 04:05:45 ok, I'll have to do some old school natting then... Jun 24 04:06:04 Well, I could install that ghetto tunelling thingee. Jun 24 04:06:07 Whatever it's called... Jun 24 04:06:35 Right, miredo. Jun 24 04:06:37 I can do that. Jun 24 04:09:21 ... waiting on sdcard ... Jun 24 04:09:33 Welcome to the life of an ARM porter. Jun 24 04:09:46 It's amazing how well you learn to multitask when you're always waiting on slow hardware. Jun 24 04:10:08 now that we have PXE in uboot I should really just make a kernel + initrd that lets me boot from nbd :) Jun 24 04:10:38 That would work. Not sure if natty would do that, though. Jun 24 04:10:40 precise should. Jun 24 04:10:47 (But I need natty today, so that doesn't help) Jun 24 04:12:30 booting Jun 24 04:21:43 infinity: getting nfs on the board and it'll be good to go Jun 24 04:21:55 stgraber: \o/ Jun 24 04:30:29 infinity: ssh ubuntu@sateda.stgraber.org -p 9922 Jun 24 04:30:45 infinity: you should have password less sudo working on it Jun 24 04:31:00 infinity: and the nfs in /data Jun 24 04:31:39 infinity: I didn't apply any update but I reboot tested the current install, so if you need you should be able to upgrade, reboot and get ssh working again :) Jun 24 04:32:23 stgraber: I'll be working in quantal chroots anyway, don't care if the base system is up-to-date, except for kernels. Jun 24 04:32:34 Though, yeah, updating the kernel to a newer natty one might happen. :) Jun 24 04:33:39 Dear NFS, 4294967294 might not be a valid UID, please to be checking your integer operations. Jun 24 04:34:07 oh, fun... Jun 24 04:37:22 infinity: looks like I was missing nfsvers=3 Jun 24 04:37:38 stgraber: Oh, should I stop what I'm doing and let you remount? Jun 24 04:37:57 infinity: yeah, I think it's best Jun 24 04:38:08 Go nuts. Jun 24 04:38:29 infinity: fixed, uids look good now Jun 24 04:38:41 And now it's sticky! Yay, giant tmp. Jun 24 04:38:49 (Not that I care) Jun 24 04:39:36 infinity: is it? it's 777 but I don't see the sticky bit here Jun 24 04:39:52 Oh, that's just ls --color confusing me. Jun 24 04:40:01 The coloring for 777 and sticky are awfully similar. Jun 24 04:40:15 (But not identical, when I look closer) Jun 24 04:44:10 stgraber: Do you have a local mirror of ports? ports.u.c seems a bit slow from your network. Jun 24 04:44:40 (I'll quit whining and just wait longer) Jun 24 04:45:47 infinity: nope, not enough ARM hardware around to justify it... though that means it's getting into squid now, so will be fast next time I want to get a quantal armel chroot ;) Jun 24 04:46:00 Heh. Jun 24 04:47:16 * infinity notes that almost every computer in his house except two relies on ports. Jun 24 04:48:11 * infinity notes that it's also a bit nerdy to say "every computer except two", as if everyone has more than two... Jun 24 04:49:09 infinity: as for speed, for some reason Canonical's network is badly routed tonight... let me see if I can fix that quickly Jun 24 04:49:35 stgraber: It's always badly-routed for me. I bounce through a GRE tunnel in San Jose to "fix" it. Jun 24 04:49:41 stgraber: Which is just hilariously silly. Jun 24 04:50:17 yeah, I'm trying to "fix" it by getting it over IPSEC to my server in Germany Jun 24 04:51:44 much better! let me quickly apply that to ports now :) Jun 24 04:51:59 Your bad routing looks less crap than my bad routing. Jun 24 04:52:37 You at least get what looks like a clear path through L3. Jun 24 04:52:54 I end up deep in some reeeeeealy old as6809.net links in the Netherlands. Jun 24 04:55:17 I guess teksavvy has some load issue with some of their IPv4 peers... The new "route" goes over IPv6 to Germany, then back over IPv4, that should workaround that problem. Jun 24 04:58:21 hmm... looks like IS hasn't investigated that routing weirdness I've had for 2 weeks now in Germany... Jun 24 04:58:50 basically, half of the Canonical subnet gets routed from Germany to the US (Boston IP) then back to London, increasing the latency quite a bit Jun 24 04:59:34 That's pretty fun... Jun 24 04:59:42 Also, I think you broke my routing mid-debootstrap. Jun 24 04:59:55 I just realised it's been downloading util-linux for the last few minutes. Jun 24 05:00:05 Oh well, squid should make a retry fast. Jun 24 05:00:11 Oh! Jun 24 05:00:29 Or you broke MY routing to YOU. Jun 24 05:00:58 Oh, no. Jun 24 05:01:01 It was the former. Jun 24 05:02:32 yeah, I guess wget doesn't like the route changing in the middle of a download Jun 24 05:02:48 No, there's actually a bug open about that. Jun 24 05:02:56 wget seems to fail miserably at timing out. Jun 24 05:03:58 infinity: routes from Germany => http://paste.ubuntu.com/1056968/ (at least it's "only" adding 80ms when going through the US) Jun 24 05:04:31 Haha. Oops. Jun 24 05:05:24 I suspect that falls firmly in the "BGP is haaaard" bucket. Jun 24 05:06:48 well, I'm not sure how Canonical's BGP is setup, but Boston should only announce the local subnet with a weight of 0 and bump the weight of the London subnets quite a bit to avoid that kind of weirdness Jun 24 05:07:00 Ideally, yes. Jun 24 05:07:29 But, for all the people who know how BGP works, it's remarkably rare to meet people who know how to implement it correctly. Jun 24 05:08:19 I consider it a minor miracle that the Internet mostly functions, most days. Jun 24 05:08:26 After having working for some T1 carriers. Jun 24 05:09:09 hehe :) Jun 24 05:09:17 Maybe things have improved since I worked for PSInet in the late 90s, but man, it was a hilarious mess back then. Jun 24 05:09:37 And the revelation that routing is pretty much a massive exercise in trust was frightening. Jun 24 05:10:14 well, I know most AS still allow all prefixes from their peers which leads to "interesting" things happening from time to time Jun 24 05:10:24 Which has led to fun foibles, like when Shaw advertised too broadly, and half of the traffic for the west coast of the US and Canada all jammed through one router in Vancouver. Jun 24 05:10:51 Turns out that was unpleasant. Jun 24 05:11:51 at least people must have noticed pretty quickly :) Jun 24 05:12:02 Oh, we noticed in a hurry. Jun 24 05:12:07 it's really easy to mess up a whole network that relys on bgp Jun 24 05:12:08 As did MCI/UU, and a few others. Jun 24 05:12:34 The guy I talked to at Shaw's NOC said the phones lit up about 30 seconds after he answered my call. Jun 24 05:14:58 This is what happens, I suppose, when you let cable companies play on the Internets. Jun 24 05:15:42 It still shocks me how quickly Shaw went from "TV operator" to "operating one of the most reliable backbones in North America". Jun 24 05:15:55 But they still mess up peering on a regular basis. Cause they're just that awesome. Jun 24 05:17:03 :) Jun 24 05:17:11 infinity: and other isp's still peer with them? Jun 24 05:17:46 gildean: They're all idiots these days, from what I can tell. Jun 24 05:17:57 gildean: They've become much more forgiving. :P Jun 24 05:18:12 and i guess if they own enough actual cable, there's little choice Jun 24 05:18:39 The halcyon days of UUnet (err, MCI, err, Worldcomm, err, whoever they are now), PSInet, and Sprint owning 99% of the North American traffic and ruling with an iron fist are LOOOONG over. Jun 24 05:19:50 Honestly, if people are still willing to peer with Cogent, that's kinda proof that they'll peer with anyone who has customers. Jun 24 05:20:16 ... he says from a server sitting on a Cogent link (it's hard to argue with cheap). Jun 24 05:22:02 and these days it's quite easy to work around isp's routing problems, just open up a ipv6-tunnel Jun 24 05:22:45 as long as you got a point of precense somewhere near, it shouldn't rise the latency more than a couple of ms Jun 24 05:23:20 and the route to the pop actually works (and doesn't get bounced around) Jun 24 05:24:08 s/and/if Jun 24 05:31:27 yeah, I quite like he.net for that ;) they have a pretty good network and a lot of points of presence so that's usually low latency too Jun 24 05:31:50 though nowadays I've got native IPv6 pretty much everywhere (most of them still using HE.net for transit though) Jun 24 06:05:33 stgraber: no native ipv6 from my isp Jun 24 06:05:39 sucks balls Jun 24 06:06:01 i use a sixxs tunnel, they had a pop righ near to me Jun 24 06:06:16 ping to pop is about 13ms Jun 24 11:59:30 <_raven> hi Jun 24 12:00:12 <_raven> do you have some experience with linux on Raspberry PI? Jun 24 12:08:34 _raven: topic Jun 24 12:26:56 is there a armhf deb floating around yet for nvidia-tegra drivers? Jun 24 14:59:26 infinity: Still need a panda? Jun 24 18:15:13 GrueMaster: Nope, stgraber hooked me up. Jun 25 00:01:24 infinity: Good to know. Didn't really want to release one from the search for ET. But I can reimage in a moment's notice if needed, and I have my local mirror. :P Jun 25 00:01:52 GrueMaster: Do you have one with a local disk? Jun 25 00:02:09 GrueMaster: stgraber's NFS solution just broke my build with make/nfs timestamp skew sadness. Jun 25 00:02:32 I have all 5 of my systems with sata usb. Jun 25 00:02:58 Maybe I should beg you for a fresh natty headless, then, and let stgraber have his diskless one back. Jun 25 00:03:04 Before I tear out more hair. Jun 25 00:03:53 But first, food. Food will make me feel better. Jun 25 00:03:57 It is faster for me to image with netboot. Installs directly to USB, and no prompting along the way. Jun 25 00:04:31 infinity: I can fairly easily switch that to a nbd export so you have raw block device access and no more nfs weirdness if that helps Jun 25 00:05:00 I'll start one now, then give you the ipv6 addr before I leave for our 24th anniversary dinner. Jun 25 00:05:34 stgraber: If that fixes timestamp skews, sure. Jun 25 00:05:45 stgraber: Or, I can steal Tobin's local-disk system. I'm not picky. Jun 25 00:05:50 But food. Jun 25 00:05:52 Back in a bit. Jun 25 00:09:27 Crap. Mirror appears to be down. Grrr. Jun 25 00:09:56 infinity: I'll have nbd setup (with the same data) in a couple of minutes Jun 25 00:14:18 Mirror recovered (kernel panic - will look into later). Jun 25 00:16:47 Of course, nothing in the logs. Oh, well. Panda4 will be online in ~15 minutes. Mirror is current as of 6/18, will pull in 45 minutes and be current within 1.5-2 hours. Jun 25 01:00:57 infinity: oh, btw, forgot to mention that the panda is back online with /data mounted over nbd, so if you still get timestamps problem, then it's a kernel bug ;) **** ENDING LOGGING AT Mon Jun 25 02:59:58 2012