Tuesday, 2017-12-19

*** NullMoogleCable <NullMoogleCable!NullMoogle@45.46.192.219> has quit IRC00:47
*** m_w <m_w!~michael@173-30-124-67.client.mchsi.com> has quit IRC01:06
*** m_w <m_w!~michael@173-30-124-67.client.mchsi.com> has joined #minnowboard01:06
*** agust <agust!~agust@p4FCB4189.dip0.t-ipconnect.de> has joined #minnowboard05:14
*** flatface <flatface!~naf@lakeport.sotnetworks.com> has joined #minnowboard08:33
*** m_w <m_w!~michael@173-30-124-67.client.mchsi.com> has quit IRC08:38
*** m_w <m_w!~michael@173-30-124-67.client.mchsi.com> has joined #minnowboard08:39
*** bluelightning <bluelightning!~paul@pdpc/supporter/professional/bluelightning> has quit IRC10:33
*** msvb-mob <msvb-mob!~michael@tmo-097-127.customers.d1-online.com> has joined #minnowboard12:27
*** msvb-mob <msvb-mob!~michael@tmo-097-127.customers.d1-online.com> has quit IRC12:52
*** msvb-mob <msvb-mob!~michael@tmo-096-3.customers.d1-online.com> has joined #minnowboard13:27
*** msvb-mob <msvb-mob!~michael@tmo-096-3.customers.d1-online.com> has quit IRC13:38
*** msvb-mob <msvb-mob!~michael@tmo-096-3.customers.d1-online.com> has joined #minnowboard13:48
*** msvb-mob <msvb-mob!~michael@tmo-096-3.customers.d1-online.com> has quit IRC13:51
*** msvb-mob <msvb-mob!~michael@tmo-096-3.customers.d1-online.com> has joined #minnowboard13:51
*** msvb-mob <msvb-mob!~michael@tmo-096-3.customers.d1-online.com> has quit IRC14:42
*** stephano <stephano!~stephano@134.134.139.72> has joined #minnowboard15:39
*** stephano <stephano!~stephano@192.55.54.41> has joined #minnowboard15:40
*** m_w <m_w!~michael@173-30-124-67.client.mchsi.com> has quit IRC17:35
furyhow fast should i be expecting the minnowboard turbot dual-ethernet's ethernet to run? i'm plugged into the left ethernet port (next to HDMI), and i'm getting 11 MB/s max transferring files to/from an M.2 SSD. running debian 9.3, ip link show says they're running at 1000 megabit18:51
furynevermind, bad cable, switch said it was only actually running at 10019:02
*** bluelightning <bluelightning!~paul@2406:e001:523f:1:5e51:4fff:febb:401d> has joined #minnowboard19:05
*** bluelightning <bluelightning!~paul@pdpc/supporter/professional/bluelightning> has joined #minnowboard19:05
*** m_w <m_w!~Michael@173-30-124-67.client.mchsi.com> has joined #minnowboard19:11
Bitweasilfury, so... what are you seeing now?  Proper gigglebit?20:35
Bitweasil11MB/s is definitely a suspicious number.20:35
fury30-40, which is the write rate i expect from the SSD - haven't googled yet how to do a raw bit rate test20:37
Bitweasiliperf should do it.20:37
fury933 mbit20:38
furysounds about right20:38
BitweasilYeah, that's sane enough for actual transfer.20:38
furyyeah. dud cable i guess - looked the same as the one i plugged in, but maybe got frayed or something and they knock it down to 100 to be safe20:39
furyboth cat5e, both blue, both 6'20:39
BitweasilOr kinked.20:40
furywell, 4'20:40
furyi've been running some benchmarks to see if an atom could really run a server. got an SSD hooked up to it, was planning to set up a kubernetes cluster, turns out 2 gigs of RAM is not enough to run a master node and anything else20:43
furyi thought i didn't need much CPU horsepower, so i was checking things out to see how it compared to the existing 7 year old Xeon server I'm currently rocking20:44
furybut when it spins all its CPU cycles on kswapd0, nothing else gets done20:44
furylol20:44
BitweasilWell, what sort of server do you need? :p20:46
BitweasilAn Atom can certainly run a server.20:46
BitweasilAn ARM chip can run a server.20:46
BitweasilBut you're not going to get it loaded down with RAM easily.20:46
BitweasilTBH, I wouldn't use a Minnowboard for an ARM server cluster.20:46
BitweasilIt's really more of a SoC breakout board.20:46
furyyes. i was thinking single board computers, x86, need at least one ethernet, maybe even two so i can split them off into a separate switch on which the cluster communication happens, that doesn't bother the main network while still getting access to them to SSH for maintenance or whatever20:47
furyboard-db.org says up squared or udoo x86, found a kubernetes cluster project on hackaday that used an udoo x86 for local development of a cluster, i'm thinking maybe just grab some of those and have at it20:47
furylattepanda delta might do the job too but it won't be out for another half a year, maybe longer20:48
furyand since i found i need more than 2 gigs of RAM per node, i might be thinking 8, top-end up squared or udoo20:48
furyand buy 5 of them and kubespray a cluster, have all of them running so that everyone has a copy of the master data and can take over if up to 2 nodes just randomly die20:49
fury5 of them is cheaper than 5 1u servers from dell, supermicro, etc20:49
furymight do the job just as well for the 5 users a day that it's currently getting20:49
fury(but i'm one of them and i do the work of like, 20 people :P)20:49
BitweasilIs this a dev server or a production box?20:50
Bitweasildev/production cluster20:50
furyi'm thinking production. current production boxes are old, one of them is rebooting (luckily, the master node >.> so everything dies)20:50
BitweasilHm.  I'd probably go with a single Xeon/ECC high reliability box.20:51
furydon't want to spend much money on a new reliable cluster, want to be able to just treat them like cattle, if one dies, push a button, flash an image or whatever, reinsert where the dead one was20:51
BitweasilYeah, I guess.20:51
BitweasilI just don't think the single board machines are the right answer there.20:51
furyi was thinking of that, and i mean, i could scale up i guess20:51
furybut if i scale up and THAT server dies20:51
BitweasilHow many 9s of uptime do you /really/ need?20:52
BitweasilNot "I would like Google level reliability on a college student's budget" - there lies madness.  Been there, done that.20:52
BitweasilI mean, your little SSD compares to a decade+ old spinning hard drive in IO.20:53
furynot expecting google level reliability :P just want less headaches, this single server setup (upon which i tacked 2 additional worker nodes to try to mitigate the effect of its failure) has given me more madness than i think 3-5 SBCs would20:53
furyyeah20:53
furymaybe i'm overthinking it, but in the other department, i've been playing with SBCs for a while and they're perfectly capable computers, don't know that i care for a 3 year warranty on it, i just want a duplicatable block that doesn't cost much, doesn't take much time to swap in, and can be added onto in case i grow past 5 users20:54
furythen again, i did just give myself a bit of a headache trying to set up a cluster on the 2 minnowboards i have - due to lack of RAM20:55
BitweasilI'd be far, far more inclined to use NUCs than Minnowboards for your use case.20:55
furyother than that it was great. ran one command line from a host machine, it spewed out a bunch of ansible stuff, set up both with docker, kubernetes, network, etc.20:55
BitweasilYou get a better CPU and can add a lot of RAM.20:55
BitweasilAnd by the time you figure in power supplies and SSD adapters and such...20:56
BitweasilThe Minnowboards have a style edge, won't argue that.20:56
furyhmm...is there a NUC that's comparable in price/performance to the minnowboard? last i checked they were all $500+20:57
furyand all had core i3 and up20:57
BitweasilThe i3s are $300ish and should blow a Minnowboard out of the water.20:57
furyi dunno, seems like the ones i can find are 4 generations old or don't come with anything. maybe i am overthinking it and/or am crazy, i just can't imagine needing much more than a smattering of SBCs on which to run the cluster rather than buy a $2-4k 1u server21:13
furynot talking raspberry pi of course, i mean, need at least 4 and prolly 8 gigs of ram just because gitlab wants a couple gigglebytes to itself, kubernetes api server might take up 2-4, etc21:14
furypotentially could just run it all on one, but want the additional 2-4 to prevent single point of failure21:15
furyeven if it is a high reliability xeon with ECC RAM, which the one it's running on is, one part could fail and the whole thing goes down until that part or that box is replaced (no website, no source control, no fun)21:15
BitweasilI think you're inventing problems you don't have, to shoehorn a solution you want...21:20
BitweasilI'm out, Git just trashed a project of mine.21:21
BitweasilI fucking hate Git.21:21
furysheeeeeit21:27
BitweasilYup.  Just lost a bunch of uncommitted changes, near as I can tell.21:29
BitweasilTrying to recover from a merge from master.21:29
BitweasilI don't understand it well enough to know how to fix it when I fuck it up.21:29
Bitweasilah, found the deleted lines.21:31
furyi've been using it for 4 years, still don't quite know myself21:32
BitweasilOk, found my work and restored it.  Still, fuck Git.21:56
*** agust <agust!~agust@p4FCB4189.dip0.t-ipconnect.de> has quit IRC22:19
*** m_w <m_w!~Michael@173-30-124-67.client.mchsi.com> has quit IRC23:20
*** m_w <m_w!~Michael@173-30-124-67.client.mchsi.com> has joined #minnowboard23:47
RjmendezBitweasil: http://ohshitgit.com/23:51
BitweasilThanks!23:58
aholleri wonder what git haters are prefering23:59
ahollercvs?23:59

Generated by irclog2html.py 2.11.0 by Marius Gedminas - find it at mg.pov.lt!