IRC Logs for #crux Wednesday, 2017-03-15

*** blueness has joined #crux00:04
*** blueness has quit IRC00:06
*** blueness has joined #crux00:07
teK_"fear"00:09
teK_pedja: a littl emore context would help00:10
*** john_cephalopoda has left #crux ("Trees are hiding until the final day.")00:10
cruxbot[opt.git/3.3]: llvm: remove stale patch00:12
pedjateK_, llvm build breaks at 'install ocaml docs' stage for some reason.ocaml bindings build and install without problems.00:13
pedjaafter I disabled them, llvm builds fine.00:15
teK_cool00:15
teK_giving it a spin00:16
pedjaRyzen 7 1700+Asus Prime X370 Pro+2x16Gb=108k RSD (~870€)00:46
*** onodera has quit IRC00:46
pedjathat's about 3 average wages (for most of the population)00:52
cruxbot[opt.git/3.3]: llvm: disable ocaml docs01:07
*** moth has quit IRC01:10
pedjathanks, teK_ :)01:21
*** estranho has quit IRC01:40
*** blueness has quit IRC01:49
*** emmett1 has quit IRC01:51
*** emmett1 has joined #crux02:28
*** blueness has joined #crux02:38
*** chinarulezzz has quit IRC02:47
*** SiFuh has quit IRC02:55
*** SiFuh has joined #crux02:57
*** chinarulezzz has joined #crux02:59
*** _________mavric6 has quit IRC03:27
*** _________mavric6 has joined #crux03:28
abenzpedja: screenies of your gorgeousness ?03:47
*** tilman_ has joined #crux05:56
*** tilman has quit IRC05:58
*** blueness has quit IRC06:21
*** blueness has joined #crux06:23
cruxbot[compat-32.git/3.3]: llvm-32: disable ocaml docs06:36
*** blueness has quit IRC06:43
*** abenz_ has joined #crux07:24
*** blueness has joined #crux08:35
*** blueness has quit IRC08:51
*** jue has joined #crux09:38
*** onodera has joined #crux09:40
*** chinarulezzz has quit IRC09:41
*** blueness has joined #crux09:58
cruxbot[opt.git/3.3]: keyutils: remove man pages conflicting with core/man-pages10:11
*** emmett1 has quit IRC10:20
*** pedja has joined #crux10:45
*** blueness has quit IRC11:34
*** blueness has joined #crux11:39
*** blueness has quit IRC11:49
*** blueness has joined #crux11:49
*** chinarulezzz has joined #crux11:58
*** blueness has quit IRC12:04
*** abenz__ has joined #crux12:05
*** onodera has quit IRC12:06
*** blueness has joined #crux12:06
*** abenz_ has quit IRC12:08
*** blueness has quit IRC12:09
*** emmett1 has joined #crux12:17
*** abenz_ has joined #crux12:22
*** abenz__ has quit IRC12:26
*** Kruppt has joined #crux12:36
*** onodera has joined #crux12:36
*** jue has quit IRC13:19
*** jue has joined #crux13:30
*** jue has quit IRC13:30
*** jue has joined #crux13:30
*** blueness has joined #crux13:56
*** jue has quit IRC14:08
*** onodera has quit IRC14:09
*** onodera has joined #crux14:53
*** blueness has quit IRC15:21
*** john_cephalopoda has joined #crux15:22
*** emmett1 has quit IRC16:08
*** emmett1 has joined #crux16:09
*** emmett1 has quit IRC16:09
*** lounge has joined #crux16:14
*** BitPuffin|osx has quit IRC16:15
*** otwieracz has quit IRC16:28
*** dlcusa has quit IRC16:28
*** dlcusa has joined #crux16:28
*** otwieracz has joined #crux16:30
*** blueness has joined #crux16:53
*** chinarulezzz has quit IRC17:29
*** BitPuffin|osx has joined #crux17:32
*** lounge has quit IRC17:36
*** blueness has quit IRC17:39
*** lounge has joined #crux17:41
*** blueness has joined #crux17:43
*** chinarulezzz has joined #crux19:48
*** lounge has quit IRC19:55
*** blueness has quit IRC19:58
*** blueness has joined #crux20:05
*** tsaop has joined #crux20:25
*** nullspoon has joined #crux20:44
nullspoonAnyone mind if I ask a non-crux programming theory related question here? I know we've got some really bright folks in this channel.20:47
*** pejman has quit IRC20:50
john_cephalopodanullspoon: Just ask :þ20:51
nullspoonJust don't want to break the topic rules.  :)20:58
nullspoonI've got something that I need a different perspective on. I've got what is effectively a collection of 4 states, which can apply to either half of a cluster, totaling 16 combinations (4^2)20:59
nullspoonEach state has a different response, though the responses don't vary much. I'm trying my best here to avoid writing just a huge set of case statements or if statements.21:00
nullspoonThe responses boil down mostly to four, so the best I can achieve is a logic tree handling the four responses. I still have to do a series of if statements that gets to one of the four responses.21:01
john_cephalopodanullspoon: What kind of responses are those?21:01
nullspoonChanging state responses. Mostly ssh commands and log messages.21:02
john_cephalopodaDifficult to say from the description alone.21:02
nullspoonfor instance... I've got a cluster. Half is live, the other is offline.21:02
nullspoonThat combination says, build the offline cluster and set it to standby21:02
nullspoonOr I have a standby cluster and the other half is offline. Set the standby cluster to live, and build the offline.21:03
nullspoonStuff like that.21:03
nullspoonI'm not entirely sure there is a better way to do this really, just a less bad way. Just thought I'd ask someone else to get at least a second pair of eyeballs on the situation. Perhaps I'm just missing something obvious.21:04
john_cephalopodaThe actions for each cluster are identical?21:04
nullspoonYeah.21:05
nullspoonWhich I think divides my responses in half.21:05
nullspoonStill, that's 821:05
nullspoonHerm. I think I can crowbar this into being a bit smaller with that.21:06
john_cephalopodaIs e.g. an offline half always built and put into standby?21:07
*** pejman has joined #crux21:07
nullspoonNot always. In the case where state determination failed on the other half, I exit with an error message indicating the user should look into it.21:08
john_cephalopodaHmmm...21:08
john_cephalopodaYou could try to make a state diagram and see if something can be made easier.21:08
nullspoonThen there's the edge case where both are offline. I only want to build one21:09
nullspoonAh, glad you asked that.  :)21:09
john_cephalopodaIf one is standby and one is offline, the standby one is turned on?21:10
nullspoonhttps://iohq.net/share/?f=V4bGCS.jpg21:10
nullspoonIf one is standby, and one is offline, deployment goes to the standby servers.21:11
nullspoonRoundabout way of saying, we always go to offline or standby servers. Live servers are never touched until they go into standby mode.21:12
cruxbot[core.git/3.3]: man-pages: updated to 4.1021:16
nullspoonThe difficult part is how we handle not having a live half.21:18
nullspoonHerm. That might have solved it actually. If I just always assume, go to the half A if no live. Else go to the other half that isn't live, regardless of its state (offline, error, or standby)21:18
john_cephalopoda[se]=sl seems wrong?21:18
john_cephalopodaWhy would the error server go live?21:19
nullspoonIt's right. Just a nuiance of the way it's handled. Live nodes are never touched, so setting a n error state to live is fine becuase it will be ignored.21:19
nullspoonI shoudl clarify, I've got a state probability function. If it errors out, then the cluster is in an inconsistent state (<51% certainty).21:20
nullspoonThat's a fine state to leave as-is (like 2 are live, 2 are offline because of an hdd crash or something) since it will still be service traffic.21:21
nullspoonserving*21:21
nullspoonHerm. I think I might be overcomplicating this.21:23
john_cephalopodanullspoon: Why 4 states?21:26
john_cephalopodaHmm, it's difficult...21:27
*** tsaop has quit IRC21:28
*** tsaop has joined #crux21:29
cruxbot[xorg.git/3.3]: libpthread-stubs: updated to 0.421:29
cruxbot[xorg.git/3.3]: xorg-server: updated to 1.19.321:29
john_cephalopodaI don't really get the whole logic behind that.21:30
john_cephalopodaTwo servers on standby -> one goes live, but one sever on standby, one off -> it stays like that.21:31
john_cephalopodaWouldn't it make more sense to always keep one server life and try to keep the other one on standby?21:31
john_cephalopoda*live21:32
*** moth has joined #crux21:33
john_cephalopodaSo for every half: "if offline -> turn to standby; if standby and half2 is live -> wait; if standby and half2 is not life -> start half number rnd%2;21:35
nullspoonThat's basically what I got to, but that's quite a few if or case statements.21:36
john_cephalopodaIf an error occurs, it enters the "error" state, which is completely ignored.21:36
nullspoonI think I can boil this down quite a bit actually, come to think of it.21:38
nullspoonIf I just check for a live state server cluster, I can skip quite a bit of the logic because the other half, regardless of state, is standby.21:38
nullspoonThen handle the edge cases where no servers are live.21:39
nullspoonReordering the state diagram to match that, that really handles a huge number of the cases in one shot.21:39
nullspoonIt actually only leaves me with two other cases - complete error states and offline/standby standby/offline (which are effectively the same)21:41
john_cephalopodanullspoon: https://bpaste.net/raw/7d48c338f5fb21:44
john_cephalopodanullspoon: I am not sure if it does what you want. It will keep one half running and freeze when one of the server errors while the other half isn't live.21:47
nullspoonThanks for writing that. It's not quite what I'm looking for, but it does give me an idea.21:50
nullspoonExactly why I asked you folks this question.  :)21:50
john_cephalopodaWhee~21:51
john_cephalopodaThe mesa3d-32 URL is wrong.22:01
Worksterseriously... goes to look22:19
john_cephalopodaIt looks like 17.0.1 isn't on that ftp (yet?)22:34
j_vftp://ftp.freedesktop.org/pub/mesa/$version/mesa-$version.tar.xz to ftp://ftp.freedesktop.org/pub/mesa/mesa-$version.tar.xz22:40
*** tsaop has quit IRC22:46

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!