00:14:53Guest69 joins
00:15:19atphoenix quits [Remote host closed the connection]
00:16:31atphoenix (atphoenix) joins
00:16:46atphoenix quits [Remote host closed the connection]
00:17:25atphoenix (atphoenix) joins
00:18:31<atphoenix>OVH and DriveSavers should have some conversations.
00:18:33<atphoenix>Data recovery from fire-damaged equipment is sometimes possible: https://drivesaversdatarecovery.com/company/museum-of-bizarre-disk-asters/
00:18:36<atphoenix>But not always: https://drivesaversdatarecovery.com/blog/fire-damaged-devices-recoverable-or-not/
00:18:38<atphoenix>OVH would also need to have details on which drive serial number was associated with which machine and customer to put things back together.
00:18:53hooway quits [Client Quit]
00:19:03Guest69 quits [Client Quit]
00:21:58<@JAA>For the cheap price of a few dozen million Euros, you, too, can get a DC worth of burnt HDDs recovered!
00:22:35<atphoenix>apply a bulk discount code to sve 25%
00:22:56<@JAA>SBG2 housed 30k servers, and each HDD costs on the order of €1-2k to recover, depending on the damage. So yeah...
00:24:04<@JAA>Not to mention that if anything of the internal structure collapsed during the fire, they might not even know which HDD belonged to which server.
00:24:09<@JAA>Also, some of them probably melted.
00:24:58<Sylirana>They *could* do it for less though. And it would definitely be some good PR for them as well.
00:25:08<atphoenix>Someone could get a lot of practice on fire-damaged-drive data recovery. Drive serial numbers should be extractable from each individual drive, assuming the drive isn't completely destroyed.
00:25:51<atphoenix>Then match drive serial to server to customer. If those records still exist.
00:25:57<@JAA>Yeah, if.
00:26:28<@JAA>And regarding PR, sure, but dropping dozens of millions into the PR department directly would probably do more.
00:27:53<atphoenix>Would insurance cover data recovery? Some policies do.
00:28:29<atphoenix>though I would tend to think barebones hosting services would skip that policy rider
00:29:02<atphoenix>where does ovh sit on the continuum of hosting providers/
00:29:06<atphoenix>?
00:30:51<atphoenix>Has anyone seen inside photos from the damaged rooms or of SBG2?
00:31:36<@JAA>I've been looking for them but haven't found any yet.
00:31:40<Sylirana>I meant for DriveSavers. This would be an opportunity for them to make OVH a space one-time offer. In return, DriveSavers will probably have more recovery tasks than they could've ever asked for and will get some good PR out of it for rescuing data in a fire that made the news worldwide. Or they can just not do nothing and get nothing in return.
00:31:59<Sylirana>*-space
00:32:11<@JAA>Ah, right.
00:34:05<atphoenix>attempting to recover, and recovering, both provide positive PR opportunities to OVH and anyone they might partner with (assuming they partner with someone who has dealt with burnt drives).
00:34:43<atphoenix>maybe some tweets are in order
00:48:27brabo quits [Remote host closed the connection]
00:51:18<atphoenix>https://www.ovhcloud.com/en/lp/status-services-backup-strasbourg/ does show that *some* SBG2 servers have recoverable 'internal backups'
00:52:37<@JAA>Well yeah, if you activated that option and ran the backups regularly.
01:21:27onetruth joins
01:42:49<Wayward->yeah, attempting to recover thousands of harddrives for the few unfortunate souls who don't know the difference between "hosting" (ephemeral) and "storage" (permanent) is a fool's errand.
01:44:46<Wayward->and I'm not sure "the cheapest dog in town" really needs to worry about PR.
01:48:27<Wayward->besides, that was an inferno. you don't recover data from drives exposed to an inferno.
01:49:36<Wayward->magnetic flux disappears at oven baking temps
02:07:31<@JAA>I doubt that.
02:10:16<@JAA>It's difficult to find hard data on this as the HDD manufacturers don't release that kind of stuff, but I'm pretty sure the Curie temperature is more like 500 °C.
02:10:34<@JAA>But yeah, no doubt some hard drives were just completely obliterated in that fire.
02:18:16<OrIdow6>That a company "should" do anything can be justified by "it would be good for PR"; the same argument can be made for why McDonald's should give away food to the poor, or why Microsoft should continue to support Windows XP
02:19:05<Wayward->Well, even the new permanent neodynium magnets out of China (not very pure) lose their magnetism at 170° C
02:19:21<Wayward->I can't imagine writable platters holding up to even that
02:20:13<@JAA>Magnetic strength doesn't have much to do with the Curie temperature.
02:21:31<@JAA>And typical 'neodymium magnets' (which aren't pure neodymium at all) have a Curie temperature of 300-400 °C.
02:21:40<Wayward->and that's not even accounting ridiculous platter expansion
02:21:52<@JAA>That's 600-750 °F for you silly units people.
02:22:19<Wayward->earlier neodynium magnets were rather pure until around 2015
02:22:25<@JAA>Uh, no.
02:22:28<Wayward->quite pure
02:22:36<Wayward->hense their expense
02:22:41<@JAA>Neodymium is non-magnetic at room temperature.
02:22:44<Wayward->now the're 1/6th the price
02:22:48<@JAA>It has a Curie temperature of 20 K or something like that.
02:23:11<OrIdow6>OVH has become popular because they have a very low standard of quality, and so can sell things at a low price; were they to do recovery like this, that image could be confused; and it could also look like an act of desparation be seen by investors as showing that they stand to lose customers if they get unlucky
02:24:10<@JAA>'Neodymium magnets' are almost always NdFeB.
02:25:00britm0b quits [Read error: Connection reset by peer]
02:25:00atphoenix quits [Remote host closed the connection]
02:25:00<@JAA>Usually Nd2Fe14B IIRC.
02:25:00sliccricc_ quits [Read error: Connection reset by peer]
02:25:00HP_Archivist quits [Remote host closed the connection]
02:25:00rsn_ joins
02:25:00qw3rty_ joins
02:25:00britm0b joins
02:25:00Atom-- joins
02:25:00sliccricc_ (sliccricc) joins
02:25:00BlueMaxima quits [Remote host closed the connection]
02:25:08atphoenix (atphoenix) joins
02:25:08HP_Archivist (HP_Archivist) joins
02:25:08BlueMaxima joins
02:25:12Atom quits [Read error: Connection reset by peer]
02:25:12<Wayward->the cheapest trick in town doesn't lose customers, it rotates customers. that's why you need to stay protected.
02:25:19Wingy quits [Quit: Ping timeout (120 seconds)]
02:25:21britmob25 quits [Quit: Ping timeout (120 seconds)]
02:25:42Wingy (Wingy) joins
02:25:44britmob25 joins
02:28:18qw3rty__ quits [Ping timeout: 264 seconds]
02:28:54rsn quits [Ping timeout: 264 seconds]
02:30:48<Wayward->JAA: i meant the neodynium *itself* used to be pure, is now no longer pure and just barely even neodynium, but a hodge podge mix of impure rare-earths with some amounts of neodynium.
02:31:35<@JAA>Yeah, that could be.
02:31:45<Wayward->that purification process is the most expensive part of rare earth magnets
02:31:58<Wayward->it's also why they've become heat sensitive
02:32:24HP_Archivist quits [Client Quit]
02:32:39HP_Archivist (HP_Archivist) joins
02:40:07<atphoenix>if you read the articles I linked to, you'll see that fire-damaged drives can be recovered, in some cases. Length of exposure, and temperatures involved, both make a difference. However, the conditions that are survivable are fairly broad.
02:59:07<Wayward->I need to find or write a tool that'll raw-read the same 4096 byte sector over and over, filling the screen with 32768 bits, so I can watch the affects of a strong magnet or blow torch in realtime
03:00:18<Wayward->alternately reading several sectors and filling the screen with ascii/ansi characters
03:00:52<atphoenix>if you figure out how to raw-read modern disks, let me know. That would be useful for data recovery via the use of statistics and/or brute force of weak bits when the CRC checks fail.
03:02:19<Wayward->i don't think there have been any attempts to prevent access to physical sector reads on platter drives
03:02:47<Wayward->any windows hex editor will display absolute sector 000000 for you and its actual contents
03:03:15<Wayward->I just don't have anything that'll auto refresh 10 times per second
03:04:47<atphoenix>there is drive firmware in between. IIRC there were some drive commands that could be used on 28-bit drives, but not 48-bit addressed drives.
03:05:25<atphoenix>that's based on vague recollections from reading hdparm and/or smartctl docs
03:05:39<Wayward->sure, firmware sectors aside. unrelated.
03:07:04<Wayward->but it would be cool to see how close I can get with a monster neodynium magnet before data becomes animated or the drive fails to read.
03:08:07<Wayward->maybe set the drive on a hot plate with a thermometer and slowly turn the dial up every 5 minutes
03:18:44<atphoenix>A few years ago I saw an interesting video, I think it was by some Sun ZFS engineers. At least is was somewhat related to ZFS, or related to one of the big rack mounted systems they were using that could take something like 50 drives. What they showed is they could *shout* at the drives and see they effects in their measurements. It might have been data transfer rates slowing down.
03:20:55<atphoenix>I think this is it: https://www.youtube.com/watch?v=tDacjrSCeq4
03:23:28<@JAA>Yeah, that one's a classic.
03:25:17Mineroboter_ joins
03:26:06onetruth quits [Read error: Connection reset by peer]
03:27:02Mineroboter quits [Ping timeout: 250 seconds]
03:27:57<atphoenix>yep. "Shouting in the Datacenter" (effects on drive performance when shouting at hard drives) aka "Brendan Gregg from Sun's Fishworks team makes an interesting discovery about inducing disk latency." with more info at http://web.archive.org/web/20090122051937/http://blogs.sun.com/brendan/entry/unusual_disk_latency
03:37:40etnguyen03 (etnguyen03) joins
04:06:02etnguyen03 quits [Client Quit]
04:24:33qw3rty__ joins
04:27:42qw3rty_ quits [Ping timeout: 250 seconds]
05:00:00treora quits [Client Quit]
05:01:15treora joins
05:04:18minus quits [Ping timeout: 264 seconds]
05:10:01minus joins
05:52:57Sylirana quits [Read error: Connection reset by peer]
05:53:56Sylirana (Sylirana) joins
06:11:32Zopolis4 quits [Remote host closed the connection]
06:48:49Zopolis4 (Zopolis4) joins
06:58:36<Wayward->videogame speed running has gotten... weird. https://youtu.be/THtbjPQFVZI
07:02:44hooway joins
07:07:58atomicthumbs quits [Quit: No Ping reply in 180 seconds.]
07:09:18atomicthumbs joins
07:20:52Doran (Doranwen) joins
07:21:06Doranwen quits [Ping timeout: 264 seconds]
07:58:37sec^nd quits [Remote host closed the connection]
07:58:49sec^nd (second) joins
08:00:18britmob25 quits [Client Quit]
08:37:31Zopolis4 quits [Remote host closed the connection]
10:16:48Arcorann (Arcorann) joins
10:53:29Arcorann_ joins
10:55:32Arcorann quits [Ping timeout: 250 seconds]
11:00:25Arcorann (Arcorann) joins
11:01:36Arcorann_ quits [Ping timeout: 250 seconds]
11:06:06Arcorann quits [Ping timeout: 264 seconds]
12:09:16BlueMaxima quits [Client Quit]
12:12:14Mateon1 quits [Ping timeout: 250 seconds]
12:14:34Mateon1 joins
12:23:56britmob25 joins
12:37:08etnguyen03 (etnguyen03) joins
14:03:40<Jake>https://twitter.com/olesovhcom/status/1372156948204949507
14:06:56@kiska quits [Quit: Ping timeout (120 seconds)]
14:08:51kiska (kiska) joins
14:08:51@ChanServ sets mode: +o kiska
14:36:06nertzy quits [Ping timeout: 250 seconds]
14:49:53rsn joins
14:52:34rsn_ quits [Ping timeout: 250 seconds]
14:52:34rsn quits [Read error: Connection reset by peer]
14:52:50rsn joins
15:04:05nertzy (nertzy) joins
15:22:45<@kiska>JAA: table etherpad plugin?
15:23:00<@kiska>Possible addition provided it works
15:24:44<@JAA>If it works and doesn't interfere with other things, why not?
15:27:49<@kiska>Doing some maintenance on pad.notkiska.pw
15:27:54<@kiska>Please do make backups
15:28:20<@JAA>I'm dumping a number of pads every few minutes.
15:28:26<@kiska>Excellent!
15:30:55<@kiska>And... something broke...
15:31:39<@kiska>And by something broke... I mean my ssh session + whatever was running since it seems to have OOM'd
15:31:45<@JAA>lol
15:32:06<@kiska>Lemme go ask Vultr for more memory :D
15:33:05<@kiska>Mem: 985M 830M 96M 1.7M 58M 55M
15:33:06<@kiska>-_-
15:34:30<@kiska>Wot...
15:34:30<@kiska>[2021-03-17 15:29:33.394] [INFO] http - HTTP server closedalled
15:34:30<@kiska> https://server8.kiska.pw/uploads/8b2d35e9c2a2eab6/image.png
15:34:59<@kiska>I am going to reboot this machine and give it some swap
15:35:12<@JAA>Let me guess, Node ate all the memory?
15:36:06<@kiska>Yes node + mysql
15:36:58<@kiska>https://server8.kiska.pw/uploads/f8026d4f9a37c92c/image.png Wonderful!
15:38:41<SCSi>oh thats nice
15:39:13<@kiska>https://server8.kiska.pw/uploads/2ebb44e9cb5617ee/image.png -_-
15:39:40<@kiska>I plan to give this 2G of swap
15:41:05<@kiska>https://server8.kiska.pw/uploads/81d2bc7d624f5134/image.png Done! :D
15:41:58<@kiska>So... this machine now has 1G RAM + 2G swap
15:42:26<SCSi>on fast drives?
15:46:09<@kiska>So... if this thing starts up, I'll go and update it
15:54:05<@HCross>kiska: want to rehome that thing in London?
15:55:21<@kiska>I've broken more things... scrypt doesn't want to install :(
16:16:59<@kiska>Lets see if this works https://www.npmjs.com/package/ep_prompt_for_name
16:21:32<@kiska>JAA: Feature installed please test if it actually works
16:45:16<@kiska>Well maintenance on the pad is over, so hopefully I don't need to look at it for the next like ever
16:57:14xit quits [Quit: Ping timeout (120 seconds)]
16:57:31xit joins
18:36:36sliccricc_ quits [Remote host closed the connection]
18:36:36mutantmonkey quits [Remote host closed the connection]
18:36:36HackMii quits [Write error: Broken pipe]
18:41:09HackMii (hacktheplanet) joins
18:42:10mutantmonkey (mutantmonkey) joins
19:06:20<@kiska>Oh hey some life coming back to SBG https://server8.kiska.pw/uploads/20131311d9ae4f16/image.png
19:07:42<@JAA>Yup: https://twitter.com/olesovhcom/status/1372210400499007491
19:08:26<@kiska>I don't quite get their visual monitoring symbols
19:08:43<@JAA>I'm glad I don't have to clean up those SBG3 servers that got smoked.
19:08:44<@kiska>Is R61 a room?
19:08:47<@JAA>12-16 hours per server... Oof
19:10:41<@JAA>I think each line there is a room. I.e. R61A, R61B, etc. are each rooms.
19:11:04<@kiska>I thought they were racks in the room?
19:11:22<@JAA>A00, A01, etc. are the racks I thought?
19:17:56<@JAA>Yeah, I think that's right.
19:21:27<@kiska>*shrug*
19:40:20DogsRNice (Webuser299) joins
19:57:53godane quits [Client Quit]
20:13:40lunik1 quits [Ping timeout: 250 seconds]
20:13:40Trieste quits [Ping timeout: 250 seconds]
20:15:03lunik1 joins
20:15:51Trieste joins
20:36:34Guest69 joins
20:56:04Guest69 quits [Client Quit]
21:11:12ROpdebee is now known as PlsNoJava
21:54:42HP_Archivist quits [Client Quit]
21:54:59HP_Archivist (HP_Archivist) joins
22:11:51BlueMaxima joins
22:47:04PlsNoJava quits [Client Quit]
22:47:27PlsNoJava joins
22:56:56godane (godane) joins
22:56:58PlsNoJava quits [Changing host]
22:56:58PlsNoJava (ROpdebee) joins
23:08:53onetruth joins
23:53:02Sylirana quits [Read error: Connection reset by peer]
23:53:11Sylirana (Sylirana) joins