tmsr - Log 2019-01-17

mircea_popescu: * shinohai has quit (Quit: No, I do not belong to The Most Delusional Republic)

shinohai: lmao forgot I had put that there when kakobrekla kept asking why I wasn't over in tmdr xD

hanbot: * shinohai has quit (Quit: No, I do not belong to The Most Delusional Republic) << the amusing part being that it'd seem that's the only republic you -do- belong to.

mircea_popescu: ouch.

asciilifeform: meanwhile uy1 cpu at 52C, raid card at 63 (for the latter, 'critical' mark is 70)

asciilifeform: still waitin' to hear from the znc-on-uy1 people, whether any want a rk acct.

asciilifeform: i can't picture why, even, unless these people baked it from ecl or sumthing

mod6: no, i mean, what's super weird is that it hasn't crashed since we started really looking into it.

asciilifeform: mod6: my current hypothesis is that it'll fall down again when raid walk.

mod6: kinda typical actually. car makes bad sound while driving it for weeks, when you bring it to the mechanic and explain the sound, he drives it, no sound.

asciilifeform: would like to , 1st, give the znc people a chance to sit down on rk, if they can be aroused to.

mircea_popescu: srsly ?

mircea_popescu: https://duckduckgo.com/?q=+%22mircea+popescu%22&t=ffsb&iax=images&ia=images << in other lulz, i fucking love the results. a buncha dorks and a buncha sluts, what more can you ask for.

mircea_popescu: in vaguely related lulz, apparently http://www.playmates.com/ died sometime last year, quietly forgotten by everyone.

mircea_popescu: all these "valuable properties" of the http://btcbase.org/log/2018-09-23#1853182 ilk, "online real estate" etc. ☝︎

a111: Logged on 2018-09-23 21:16 mircea_popescu: which is eerily similar to what these shits ever sell on an open market, be it myspace, slashdot, or whatever else.

mircea_popescu: https://www.playboy.com/read/don-t-fall-for-the-gop-s-scapegoating-of-steve-king-1 << domain squatted by Really Angry Pantsuit (tm), amusingly enough. because totally, that's exactly how this game works, http://btcbase.org/log/2015-08-21#1247708 etcetera. ☝︎

a111: Logged on 2015-08-21 23:21 mircea_popescu: "Welcome, kiddo, to how the real world works. You go ahead and place your crazy protest bids way off in the bottom of the sociological order book and scream that someday, SOMEDAY, the market will come down to meet you, and you will be vindicated.

mircea_popescu: we're totally gonna start caring about pantsuit talking points nao!

mircea_popescu: "Jeffrey Smith ‏Verified account @Smith_JeffreyT 3h3 hours ago Jeffrey Smith Retweeted Jon Ossoff

mircea_popescu: The apparent murder of journalist Ahmed Divela is chilling. Ghana is recognized by many, including me, as a vibrant, strong democracy and an example for the continent in many ways. " meanwhile in related lulz. ☟︎☟︎

mircea_popescu: * deedbot has quit (*.net *.split) <

asciilifeform: meanwhile in iron noose, https://www.chip1stop.com/USA/en/view/dispDetail/DispDetail?partId=SOCI-0000001&cid=SOCIEB << chinese, claims to sell 24core arm box with expandable ecc ram holes. ~1.2k usd.

asciilifeform: ( whether subj item is terraformable rk-style, is yet unknown to asciilifeform )

mircea_popescu: https://metapedia.org/ << in other lulz, i had nfi item even existed. anyone know what "traditional enemies" even means there ?!

asciilifeform: https://archive.is/9NDzd << cached. still can't make heads or tails of it tho.

mircea_popescu: so like wiki-*

asciilifeform: if it was somehow diff from ordinary pediwik, this cannot be determined from the cache

mod6: Would be awesome if Pizzaro could sell iron that can run cuntoo + ffa + fg + trb out of the gate [ + others too, i.e. eulora, etc ]

mod6: If that chinese iron could get the job done for a reasonable rate.. could be pretty good offering.

asciilifeform: mod6: currently much moar pedestrian concern: seems like we have 1st instance of a dead rk.

asciilifeform: mod6: i'ma ship new rk + a few spares to BingoBoingo ( who appears to have solved his orc customs problems ) , but the find is concerning in re the wisdom of installing 9000 of'em

asciilifeform: BingoBoingo agreed to house lobbes et al on his personal rk, until the swap is made, he will fill in the details when he gets back to his console

asciilifeform: mod6: note that there is still not afaik anything like a candidate for 'rk, but better'

asciilifeform: it appears that we'll still have to construct a larger rk plant, but they will have to be physically racked in such a way as to be replaceable without taking down whole plant

asciilifeform looked at erry available rk board on market in last coupla hrs, they are all either of substantially inferior spec, or cost no less than 3x moar per

mod6: We aught to look at costs of full size env's that would fit the bill for TMSR~ tooling. Then establish how many we would need & charge for to get out of the red.

mod6: Envs that have at least 4 logical cores, 8 Gb RAM, and minimum 1Tb ssd.

asciilifeform: mod6: afaik that means 1u x86, given current state of art.

mod6: I guess I should say, that I'm not sure exactly. But we should do a bunch of anaysis to see what would be the best play for capital allocation.

asciilifeform: and these have monumental cost of transport, as already established.

asciilifeform: is the reason why asciilifeform went with 'pack as much into ea. 1u x86 as physically possible' when init buildout.

mircea_popescu: http://archive.is/3cRY7#selection-1855.167-1855.241 << in other lulz, notwithstanding the text, there depicted is ~romanian~ (if the soviet-style ro army uniforms didn't make it plain enough) nicolae ciuca.

mircea_popescu: !#s ciuca

mircea_popescu: hmmm... i thought i mentioned the fellow.

mircea_popescu: am i oversensitive or are the dorks actually trying to pass the guy for guler, on the grounds that "what do ustards know" ?

mircea_popescu: asciilifeform what was the rk problem anyways ?

asciilifeform: mircea_popescu: i strongly suspect that yer right, and the monkeys went 'eh, orcs, what's 1, what's another'

asciilifeform: mircea_popescu: the idle rk that was to house lobbes et al, has turned up dead.

mircea_popescu: as in won't boot ?!

asciilifeform: cpu executes random instruction periodically, with lulzy results

mircea_popescu: sounds like a botched fw "upgrade" ?

mircea_popescu: how do we know ?

asciilifeform: it runs instrs from mask rom on boot, then loads my kernel.

mircea_popescu: i have ~never~ heard of a cpu that ~genuinely~ fandango'd instructions. each and EVERY single one of the ~scant few~ dozen cases was stuxnet-like

asciilifeform: i'ma have BingoBoingo send me the retired board via snailmails, possibly we find sumething interesting about it

mircea_popescu: how the fuck would this even work, indulge me. arm processor just invents rng opcode ?

asciilifeform: mircea_popescu: i'ma post lulzdump for the pleasure of mircea_popescu et al in a min.

mircea_popescu: cool.

mircea_popescu pages ave1 too, it'd be a pity to miss such

asciilifeform: and in between these actually booted up and ssh-reachable but then panicked in few min

asciilifeform: all 4 .txt are from the serial console uart, if it aint obvious

asciilifeform: initially suspected a dead drive, these produce similar symptoms ( garbage when reading disk ), had BingoBoingo fetch brand-new one from spares stash

asciilifeform: near as i can tell, thing ends up fetching bitrotted blox from disk (ssd stix) and dies, unsurprisingly, at ~rando

asciilifeform: near as i can tell, thing ends up fetching bitrotted blox from disk (ssd stix) and dies, unsurprisingly, at ~random points in first few min of life

mircea_popescu: so the init doesn't complete

asciilifeform: mod6: BingoBoingo added forced air when we installed disks

asciilifeform: mircea_popescu: on 1 run, completed, death came 30sec into ssh session

asciilifeform: mod6: that's neither here nor there, it's a cpu temp reader thing

asciilifeform: asciilifeform's kernel for rk does not support loadable modules

mircea_popescu: looky : yet in ~every~ case it initializes the ram.

asciilifeform: asciilifeform-baked softs on this unit are identical to the other 5.

asciilifeform: mircea_popescu: not only, it does make it to the kernel

mircea_popescu: that's by far the larger spot. how does this broken cpu get through ddr ?

asciilifeform: mircea_popescu: i suspect zapped , somehow, usb controller

asciilifeform: possibly as consequence of cleaning, or of the earlier dud drives ( tho none of other 5 were similarly afflicted )

asciilifeform: i.e. it gets through 1) mask rom 2) load of kernel from sd

asciilifeform: hence why i initially said 'hey, BingoBoingo , loox like another dead stick..'

asciilifeform: notably, a test using the auxiliary (usb2.0) jack returned same result.

asciilifeform: if mircea_popescu or anybody else can think of other interesting tests to conduct on this sad box, i'm all ears.

asciilifeform: a test with independent (from other 5) power supply, also returned same.

asciilifeform: BingoBoingo sweated in the cage for last coupla hrs , tryin' these.

asciilifeform: i asked him to go there when i sat down to make accts for the uy1 people on the thing, and found that : 'ssh_exchange_identification: read: Connection reset by peer'

asciilifeform: previously only ever saw this on boxes with dead disks. what happens there is that sshd eats new connection and then goes to read from hdd and unix eggogs, so it drops

asciilifeform: on reset , thing failed to revv up to pingability at all. so i asked for the uart snake.

asciilifeform: i picked rk orig. not even because of price, but specifically cuz it contains, on the pcb itself, nuffin (known to be) writeable, all of state is loaded from sd and afterwards usb. so i dun expect to be able to tell much from autopsy of the unit, when i get it here

asciilifeform: given as the other 5 units of pilot plant are still standing, i suspect straight chinese rot, rather than bomb.

mircea_popescu: well, i'm not sure what your decision tree is like, exactly. but proceeding logically : 1) it can not be the case cpu is sad on rk ; 2) it is not shown that deffect is organic rather than nsatronic ; 3) the selling point of rk is cheapness/disposability. you can afford to throw these away, unlike an adult server board+deckings.

mircea_popescu: now, if indeed per 2 enemy can perma-trash your boxes, then 3 becomes moot.

asciilifeform: not sure why enemy would start with idle unit rather than, e.g., diana_coman's, tho

mircea_popescu: otherwise, you amply do not care : a box that crashes on diddling is as valuable as an undiddlable box.

mircea_popescu: they are affordable, and so you have a very strong position there.

asciilifeform: mircea_popescu: hence my earlier observation. we still gotta build 2nd plant, but it'll have to be constructed slightly differently, initially i was gonna stack'em

mircea_popescu: asciilifeform im just saying, that the only serious problem here is if in fact eve can fuck your box by feeding it magic packet

asciilifeform: mircea_popescu: if 'magic packet', then decision tree of enemy is somewhat perplexing, why start with only 1 box ( why not nuke all 6 ? and have asciilifeform chasing dud power supply etc ) , or why not start with a live box.

mircea_popescu: nfi.

mircea_popescu: it certainly could be a failing coupling in usb jack, or somesuch.

asciilifeform: the problem with the whole thing being 'disposable', i.e. housed on 3chips (the cpu, ethernet, and ram, seem to be the only separate crystals on the pcb) , is that nuffin useful can be squeezed from the dud.

asciilifeform: if one supposes 'organic' failure, 1 out of 6 boards smoked in the span of <9 months, is pretty high smoke rate.

asciilifeform: i'd quite like to 'ha, i'ma throw rk, and use $x', but i haven't any lead on a usable $x.

BingoBoingo suspects USB controller gone sad, Going to extract sad RK when the fingers are feeling nimbler

asciilifeform: BingoBoingo: quite annoyingly, usb controller is on-die peripheral on rk3328.

asciilifeform: so not even possible to 'let's detach from pcb in asciilifeform's torture room and see if dud usb ic'

asciilifeform: mod6: why didja think they were small & cheap. nearly errything is on 1 ic.

mircea_popescu: i still don't see the problem, 1 in 6 failure rate ? was the one closest to the suspected hotbox from prev week ?

asciilifeform: mircea_popescu: they're all in a breadbox-sized shelf, quite far from uy1. and with own forced air.

mircea_popescu: logs yu published do look like intermittend ssd failure.

BingoBoingo: mircea_popescu: It's not particularly close. Spent a good chunk of time cleaning it yesterday. This particular one has been iffy since the sad cheep USBs started failing.

mircea_popescu: yes.

mircea_popescu: BingoBoingo what;s iffy mean here ?

BingoBoingo: mircea_popescu: Took multiple power cycling to boot up in a way that responds to ping.

asciilifeform: BingoBoingo: iirc it stood up when we swapped disks the earlier time. but no one's lived in it since.

diana_coman: did this RK died while being unused ? a log search for it shows only something in August last year - new master img deployed http://btcbase.org/log/2018-08-08#1840376 ☝︎

a111: Logged on 2018-08-08 16:58 asciilifeform: BingoBoingo: i'ma test new kernel today, then tomorrow will ask BingoBoingo to power down rk-c and put disk in dulap to i can snap new master img, then it will go into service ( and can reimage BingoBoingo also ) .

asciilifeform: uninhabited since it disk swap; last housed mod6 ( who moved to a vacant, at the time, unit, when we swapped disks )

BingoBoingo: <asciilifeform> uninhabited since it disk swap; last housed mod6 ( who moved to a vacant, at the time, unit, when we swapped disks ) << Uninhabited last housed me.

asciilifeform: diana_coman's point stands tho, conceivable that it was already rotten even then

BingoBoingo: Is that vacant one because stood up with the vacant RK drive but fought my image.

diana_coman: is that repeated power cycle something that was encountered before?

asciilifeform: diana_coman: not iirc. BingoBoingo did you observe this prior ?

diana_coman: or was it even possible to figure out what required it / what was the underlying issue?

asciilifeform: diana_coman: the problem with these boxes, is that there is essentially 0 meaningful debug possible

BingoBoingo: asciilifeform: It's the behavior we observed with the sad drives. Have not observed this post sad drives on any other RK.

diana_coman: yes, I'm scratching here at anything at all; not much I can see otherwise

asciilifeform: diana_coman: i'm all ears, even re the oddest hypothesis anybody can think of.

diana_coman: did this one die with a sad drive too or was it changed before it really died on it?

asciilifeform: was stood up again with new drive. ( and then ~again~ today with fresh one , of same type as presently used by e.g. diana_coman ), same res.

diana_coman: so at first new drive it required several power cycles but then it seemed fine?

BingoBoingo: residue from sad drive contacts was was removed like on the other RK's

asciilifeform: asciilifeform was gettin' ready to grumble re rotten usb drives in the l0gz today. instead nao grumbles re dead rk..

BingoBoingo: <diana_coman> so at first new drive it required several power cycles but then it seemed fine? << But only with the vacant drive. Not with my RK drive.

asciilifeform: BingoBoingo: i think diana_coman's q is re the time it was stood up with new drive in aug. do you recall any peculiarity re rk-c at end of that ?

diana_coman: I guess it's hard to tell whether it wasn't rotten at that time already; for some reason (what?!) worse than the others aka not just the drive

asciilifeform: diana_coman: i'd luvv to establish an exact time of death for the thing. presently unknown tho.

diana_coman: I suppose it might be worth adding some automated daily check at least connects? gives time of day/whatever? for unused units

diana_coman: so at least you don't get surprises when you need to use them

asciilifeform: diana_coman: once we actually have vacant rk again, i'ma absolutely institute mechanical life tests for'em.

BingoBoingo: It had more drive insertions, removals, and power cycles than the others.

BingoBoingo: FWIW the email spam folder has a lot of spam with attachments today. Usual month ~0 of these

mircea_popescu: http://trilema.com/2010/cetateni-jurnalisti-si-liceeni-jurnalisti/#selection-77.445-77.1377 << in other lulz, check out the endless list of names of anonymous nobodies.

mircea_popescu: short decade ago, these self-same morons convinced themselves and maybe any dependent preteen children / some dumber than average wives&girlfriends they "run the world", or as close an approximation to that vaunted state as practicable.

mircea_popescu: remarkably no jews there either.

BingoBoingo: If you have been using Pizarro shared hosting your IRC bouncer you are invited to stand up an IRC bouncer on my Rockchip as the vacant Rockchip appears to have failed.

lobbes: ty BingoBoingo. login successful. Will attempt irssi setup later and report back ☟︎