Show Idle (> d.) Chans
Results 1 ... 128 found in asciilifeform for 'wedge'
vex: it'd get wedged between the first two trees it saw tho
asciilifeform: cgra: would depend on what means 'plugged shut' neh. atm w/ wedge bugs, the world-facing noad (and sumbody has to have'em, or no blox) can be wedged
asciilifeform: billymg: seems like his wedged properly, rather than simply slow
asciilifeform thus far satisfied that 'cement' behaves logically correctly (i.e. 'makecement' in fact emits the longest chain, 'loadcement' in fact loads, 'unloadcement' -- unloads, and noad w/ cement loaded does not get 'wedged' for any appreciable length of time.) but defo worth extensive test by other folx.
dulapbot: Logged on 2022-02-18 08:46:58 cgra: http://logs.nosuchlabs.com/log/asciilifeform/2022-02-17#1079773 << mech hdd with the bdbmay prove unfeasible, while still an interesting experiment
cgra: http://logs.nosuchlabs.com/log/asciilifeform/2022-02-17#1079773 << mech hdd with the bdbmay prove unfeasible, while still an interesting experiment
dulapbot: spyked last seen in #asciilifeform on 2020-06-18 10:08:47: asciilifeform, danke. fixed and answered: http://thetarpit.org/2020/a-cursory-look-at-the-infamous-trb-wedge-bug#comment-353
dulapbot: Logged on 2022-01-11 18:29:12 asciilifeform: (iirc at least 1 of cgra's findings was a wedge w/out necessarily oom death)
asciilifeform: (iirc at least 1 of cgra's findings was a wedge w/out necessarily oom death)
asciilifeform: (wedge detector)
asciilifeform: thimbronion: my blatta's been wedged since after this line
asciilifeform: imho it doesn't make sense to put sweat in it until there's a trb that doesn't verify in o(n^2) and doesn't wedge, lol
cgra: i mean, similar self-shitting as the 'getdata wedge' does
dulapbot: Logged on 2021-11-16 02:43:38 vex: this is generally considered common knowedge
vex: this is generally considered common knowedge
asciilifeform: (well that an' 'wedger.py, lol)
asciilifeform not yet observed in the wild a noad wedged by being fed a constructed >pow fauxchain. dun mean it's impossible.
asciilifeform: cgra: in practice, is ~impossible to use 'cooked' faux-old blox to perma-wedge a node -- the victim would have to be connected to the perp and no one else. but they do waste substantial cpu time.
asciilifeform: mod6: cgra discovered coupla moar potential wedges, he's in the process of writing'em up
mod6: Also, I should note re-wedge 2020, I was running with the experimental patch at the time; my nodes didn't miss a beat.
asciilifeform: ( most recently, was instrumental in resolving the 2020 wedge enigma. )
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2021-09-05#1056505 << rright, loos like signpost already said. btw i implemented 'who-gave' specifically so that folx could curate their heathen-edge connections. asciilifeform did, at one time, then noticed that it made 0 diff , trb noades do not fall behind because poor edge graph, but because remote-wedgeability bugs, or net downtime, and even small 'behind' is difficult
asciilifeform: shinohai: this is only useful on a box where no circular wedges
asciilifeform: cgra: the other possibly relevant detail is that it wasn't a quick-oom , but a wedge where node very slowly shits itself while unresponsive to commands, and at very little expense from trigger end.
asciilifeform: cgra: iirc that particular bug was described in a prb communique. so, possible that the 'wedger' simply did not know of others.
dulapbot: Logged on 2021-08-12 09:13:49 cgra: fwiw, judging by the amount of OOM holes in trb, and absence of other issues similar to the 'wedge', seems to me the wedge was just someone pulling blocks in maximum force, for whatever reason, unaware of the lethal side effects. all these added up, obey_sendbuffersize may have been better off not rating such peers for misbehavior
cgra: fwiw, judging by the amount of OOM holes in trb, and absence of other issues similar to the 'wedge', seems to me the wedge was just someone pulling blocks in maximum force, for whatever reason, unaware of the lethal side effects. all these added up, obey_sendbuffersize may have been better off not rating such peers for misbehavior
asciilifeform: (the previous wedger for instance did not)
asciilifeform: if it's a troo leak, wedger-style, oughta be able to fill it neh
dulapbot: Logged on 2020-02-26 16:13:05 asciilifeform: bvt: http://nosuchlabs.com/pub/wedge/wedger.py ; and use with http://nosuchlabs.com/pub/wedge/snap_49999.txt or hash file of same format of your choice.
billymg: asciilifeform: why not simply 'import wedger' at top of watchglass then?
dulapbot: Logged on 2020-02-26 16:13:05 asciilifeform: bvt: http://nosuchlabs.com/pub/wedge/wedger.py ; and use with http://nosuchlabs.com/pub/wedge/snap_49999.txt or hash file of same format of your choice.
asciilifeform: billymg: in 'wedger.py' already extracted.
dulapbot: spyked last seen in #asciilifeform on 2020-06-18 10:08:47: asciilifeform, danke. fixed and answered: http://thetarpit.org/2020/a-cursory-look-at-the-infamous-trb-wedge-bug#comment-353
dulapbot: Logged on 2020-02-26 16:13:05 asciilifeform: bvt: http://nosuchlabs.com/pub/wedge/wedger.py ; and use with http://nosuchlabs.com/pub/wedge/snap_49999.txt or hash file of same format of your choice.
snsabot: Logged on 2020-02-26 16:13:05 asciilifeform: bvt: http://nosuchlabs.com/pub/wedge/wedger.py ; and use with http://nosuchlabs.com/pub/wedge/snap_49999.txt or hash file of same format of your choice.
asciilifeform: i've been able to keep it going (i.e. eats fresh blox from planet, sends/receives coinz) w/ trivial changes (most recently, the wedge fix w/ aid of mod6)
asciilifeform: aaan we're back. loox like 2nd time since 1st prototype that bot wedged in socket limbo.
adlai: maybe they're all hosted on the same computer, and the only melanoma is a quazicrystal of cloudflare wedged in between that and thou.
lobbes: http://alt.ericbenevides.com/log/asciilifeform/2020-11-26#1025362 << I'm pretty bad at it still, but even with my poor technique I am finding it a very useful tool for clearing the overgrown bits wedged in awkward corners and such. It is also surprisingly easy to adjust for uneven ground and for avoiding rocks and other objects (which is good since there's a type of tree here that litters my yard
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2020-09-26#1022499 << ty trinque . seems to work nao. ( outta curiosity, any idea why wedged ? was ircism ? )
adlai: thank you, trinque ; as you probably figured out, "still" was due to uncertainty wrt how much of deedbot was wedged.
asciilifeform: shinohai: pretty sure it's simply wedged atm
asciilifeform: hm trinque is deedbot wedged perchance ?
snsabot: spyked last seen in #asciilifeform on 2020-06-18 10:08:47: asciilifeform, danke. fixed and answered: http://thetarpit.org/2020/a-cursory-look-at-the-infamous-trb-wedge-bug#comment-353
asciilifeform: jurov ^ i suspect yers is wedged
snsabot: spyked last seen in #asciilifeform on 2020-06-18 10:08:47: asciilifeform, danke. fixed and answered: http://thetarpit.org/2020/a-cursory-look-at-the-infamous-trb-wedge-bug#comment-353
asciilifeform: 'people' who are firmly wedged into chickenlike flock behaviour mode aint teachable. even if somehow on surface 'willing'.
snsabot: spyked last seen in #asciilifeform on 2020-06-18 10:08:47: asciilifeform, danke. fixed and answered: http://thetarpit.org/2020/a-cursory-look-at-the-infamous-trb-wedge-bug#comment-353
feedbot: http://thetarpit.org/2020/a-cursory-look-at-the-infamous-trb-wedge-bug << The Tar Pit -- A cursory look at the infamous TRB "wedge" bug
BingoBoingo: joerodgers: The instructions one the site are incredibly on point. I recently rebuilt with the latest anti-wedge patch. A++
asciilifeform: the interesting bit is that i had the timeout in the genesis version, and it in fact didn't help. because of fleanode's peculiar zombie effect. the detector was required to get rid of all wedges.
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2020-05-06#1011292 << this was the final solution to bot wedges, rather than that
asciilifeform: wedged ? for 7d nao ? ^
snsabot: Logged on 2020-02-26 16:13:05 asciilifeform: bvt: http://nosuchlabs.com/pub/wedge/wedger.py ; and use with http://nosuchlabs.com/pub/wedge/snap_49999.txt or hash file of same format of your choice.
asciilifeform: interesting, .27 almost 200blox behind despite showing 0 signs of wedgeism. simply as if no one sent it blox.
asciilifeform: pretty interesting irc bugola : fleanode box 204.225.96.251 barfed in such a way that bot wedged w/out detecting disconnect
asciilifeform: ^ mod6's post-mortem re the wedge episode. rec'd reading for all trb operators.
asciilifeform: would be interesting to learn whether they could in principle wedge a 'cold start' trb that -connect's to one
asciilifeform: it wasn't wedged tho. someone gave it a clean shutdown
asciilifeform: 'received block 0000000000000000000e' << evidently dun have the hash untruncation patch, not to mention unwedger?
asciilifeform: ( 'wedge' dun normally give 100% stone-cold 'connection refused' )
asciilifeform: prb imho has exactly 0 to teach. even my very short look into various prb vers during wedge detective work, only mislead and delayed
bvt: yes, this i got -- when i was asking for a wedge, i had a bit patched up trb that used more proper types (64bit) in CNode net code and CDataStream -- there it allocated 5.4Gb, threw exception on next resize (raising question about what else can throw in "inconvenient" places), got caught in another infinite loop.
asciilifeform: bvt: what happens in the unpatched trb on 'wedge test', is that it actually queues up a 4+GB outgoing msg. overflows counter (32bit) .
asciilifeform: imho it is rude to wedge people, so i'ma leave off doing so (unless requested.) but 'wedger' was posted yest. and others may.
asciilifeform: confirmed 'getdata'-wedge-proof .
asciilifeform ftr didn't feed'em 'wedge'
asciilifeform: wedger indeed worx... ( ./wedger.py 205.134.172.4 8333 snap_49999.txt )
asciilifeform: fwiw it didn't perma-wedge my noad. may require looped bombing
asciilifeform: bvt: http://nosuchlabs.com/pub/wedge/wedger.py ; and use with http://nosuchlabs.com/pub/wedge/snap_49999.txt or hash file of same format of your choice.
asciilifeform: re wedge -- given that apparently 0 prb nodes suffered the wedge, i'm inclined to think that this is the answer. even tho atm cannot readily say how to replicate condition.
asciilifeform: ( this is even supposing 'wedge packet' is 'natural', i.e. accidental, rather than crafted, object )
asciilifeform: so doubtful re wedge conditions
Vexual: yeah, I aslo kinda assumed it might be useful for repeating wedge conditions
asciilifeform: imho ideally it'd probe in background (say, erry 20min) and report strictly wedged and 'behind' nodez
asciilifeform: of 'bitcoind-wedge2' : 793e8fe8d37620b6f9b6cb14e9e40e04cb45f1e9bce3ff3785a3ed427c455f1442bd35b17e1fd6746a6e321154ea5c81ab439e545dcd32a91cfcad52bb38bb02 . )
bvt: nothing interesting http://bvt-trace.net/2020/02/a-tiny-and-incomplete-trb-wedgetrace/#comment-135 (I'll use comments because pastes tend to disappear)
bvt: i have it in debugger in wedged state right now, so i'd like to get a maximum value out of it before killing.
asciilifeform: bvt: i dun suppose you have a core dump of wedged state ?
asciilifeform: sumbody's wedged...
feedbot: http://bvt-trace.net/2020/02/a-tiny-and-incomplete-trb-wedgetrace/ << bvt's backtrace -- A tiny and incomplete TRB wedgetrace
asciilifeform: ( converse not troo : if not responds -- not necessarily wedged; could be e.g. verifying heavy block at the time )
asciilifeform: re 'wedges' -- if yer noad responds to watchglass , it aint wedged.
snsabot: Logged on 2020-02-22 22:55:40 asciilifeform: BingoBoingo: 'great' job 'trb nodes have been wedged' w/out reporting who clued you in or linking to thrd. ( maybe next time will see how long it takes to notice, instead of giving what to crib? ) and et tu, mod6
asciilifeform: http://logs.nosuchlabs.com/log/asciilifeform/2020-02-23#1007461 << ty mod6 . this will come in handy when you, or i, or some hypothetical but not yet visible third party finds the time to put in the sweat to actually bake 'replay' system -- and hence make wedges reproducible.
asciilifeform: mod6: i'm satisfied that you saw proper wedge. ( fwiw, mine are atm @ 618665 and 618607 , where the lagging 1 is on the slower residential pipe. neither wedged presently. )
asciilifeform: that, there, is wedge.
asciilifeform: mod6: if you aint wedged -- then if aint getting blox, it is because other folx aint sending'em.
asciilifeform: we used to do this regularly, mod6 , you'd see a wedge and 'read clouds' , forgetting that the log ~always~ contains large amt of very similar liquishit
snsabot: (ossasepia) 2020-02-22 BingoBoingo: In other news, a number of visible trb nodes have been wedged for some time.
asciilifeform: BingoBoingo: 'great' job 'trb nodes have been wedged' w/out reporting who clued you in or linking to thrd. ( maybe next time will see how long it takes to notice, instead of giving what to crib? ) and et tu, mod6
asciilifeform: mod6: has it occured to you to compare to usual debug output ? you aint wedged. not getting blocks as fast as you like cuz yer connected to a bunch of folx who ~are~. ( if yer node accepts e.g. getinfo cmd, you aint wedged )
asciilifeform: atm much moar interested in e.g. potential race wedge in trb, then in what happens in vatican
asciilifeform: mod6: atm both of mine at 618417. not wedged, but also not yet seen next block from that.
asciilifeform: BingoBoingo: be sure to save the log before resetting any of your wedged ones, also
BingoBoingo: asciilifeform: Most of mine wedged, yes
asciilifeform: BingoBoingo yer still wedged, btw
asciilifeform: when unwedged it, that is
asciilifeform: other folx appear to remain in wedge ( trinque et al )
mod6: hm. see I don't think it has anything to do with the block. which is why I try to stay away from 'wedged', unlike the other circumstances where we've said 'wedged'. i forget the block numbers off hand.
asciilifeform: folx try to connect to wedged node, you get a 'accepted connection x.x.x.x:x'; then socket no message in first 60 seconds, 1 0
asciilifeform: ( and not necessarily inclusive of the 1 who served you the hypothetical wedge cmd . that 1 could have long disconnected )
asciilifeform: yer seeing whatever peers you had at the time of wedge
asciilifeform: mod6: the ips visible in the log aint in any way interesting, necessarily. you get same picture (folx disconnect) whenever verification aint working (because that thread wedged)
asciilifeform: mod6: when was last time you saw >12h wedge ?
asciilifeform: wedge took place immed. prior to 618406 .
asciilifeform: mod6: i saw last night, decided to hold off ringing the bell until wake up (if still wedged)
asciilifeform: mod6: i dun recall seeing hard wedge on my nodez since early '18
asciilifeform: last thought about this 2y ago, but afaik the only practical means to actually debug hypothetical 'magic mempool turd that wedges' would be to keep a ring buffer of all traffic (can reset when successful block eat, erry time) and dump to disk on sigkill
asciilifeform: ACHTUNG trb operators! loox like all trb nodes known to me are presently wedged (at 618405)
asciilifeform: today thing runs (at least, mine, but from appearances all of the public nodes) w/out handholding. haven't noticed a day+ wedge since 'aggression' on mine.
asciilifeform: today, if you 'sync', you will get bricked 'portage'. i.e. wedged because ~all~ ports now req. poetteringism to be enabled. (and gcc 5+ .)
snsabot: Logged on 2019-12-30 17:07:28 asciilifeform: shinohai: realized the need for this while attempting this earlier item -- apparently gpg 1.4 has known (but afaik not fixed) cpu wedge attacks via crafted input turds (and some unknown # of unknown... massive ball o'shit) and really not suitable for 'anyone can feed' www system
asciilifeform: shinohai: realized the need for this while attempting this earlier item -- apparently gpg 1.4 has known (but afaik not fixed) cpu wedge attacks via crafted input turds (and some unknown # of unknown... massive ball o'shit) and really not suitable for 'anyone can feed' www system
asciilifeform: waiting for repair d00d to return w / batteries for the locker thing ( ancient batts were wedged so thoroughly , fella struggled, finally i plucked'em out w/ own knife )
BingoBoingo: Surprisinly no epoxy or glue in the construction. Small plastic spring/wedge is all that secure the working part to the shell
asciilifeform: mod6: ftr i in fact owe him 0.3 ( expect to settle today, finally figured out why my wallet refill tx was wedged, it somehow was set to 0 fee. )
Random(asciilifeform) | Download hourly DB snapshot | Get Source Code