July 11, 2012 
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | >
July 11, 2012 [00:01:13] *** denizr has quit IRC[00:01:35] *** denizr has joined #smartos[00:04:27] *** joedevivo has quit IRC[00:07:57] <reiddraper> I'm trying to follow the instructions here: http://wiki.smartos.org/display/DOC/Managing+Datasets,[00:08:28] <reiddraper> but I don't have a zones/$UUID-disk0 listing, just zones/$UUID and zones/$UUID/cores[00:09:42] <reiddraper> any ideas?[00:10:06] <reiddraper> (the VM has been previously started, and is currently stopped)[00:17:28] <nahamu> reiddraper: did you use zfs list?[00:17:50] <reiddraper> nahamu: yes[00:18:05] <nahamu> hrm. what did you use as your base image?[00:19:22] <reiddraper> "dataset_uuid":"f9e4be48-9466-11e1-bc41-9f993f5dff36"[00:20:46] <nahamu> ah, that's joyent branded zone, not a KVM branded one.[00:21:15] <nahamu> yeah, it's a bit trickier to create your own template out of those, and that wiki page doesn't document it.[00:21:43] <nahamu> a crude approximation would be to clone the zones/$UUID and use that as the basis[00:21:56] <reiddraper> ah ok, cool.[00:22:05] <nahamu> but it will be missing the zones/f9e4be48-9466-11e1-bc41-9f993f5dff36/root/root/zoneinit* stuff[00:22:13] <nahamu> (which you could potentially copy in)[00:22:54] <nahamu> I think the person I'd point you at is traveling currently (he's certainly not in here at the moment)[00:23:41] <nahamu> basically, if you just clone your existing zone, the lack of zoneinit might confuse vmadm a bit.[00:23:45] <nahamu> but it should mostly work[00:24:07] <nahamu> and you can experiment with copying it back in.[00:24:29] <nahamu> this is something I should be working on in the next few days, so if I get a chance, I'll try to remember to put some notes into the wiki[00:24:31] <reiddraper> ok, so is cloning smartOS zones sort of 2nd class to KVM?[00:24:56] <nahamu> it's just not documented in the wiki and it's slightly complicated by the zoneinit stuff.[00:25:03] <reiddraper> ok, got it[00:25:12] <reiddraper> nahamu: appreciate your help[00:26:04] <nahamu> no problem. good luck! also, if you figure out the zoneinit stuff, feel free to sign up for a account on the wiki and add your notes in there. :)[00:26:19] <reiddraper> will do[00:28:49] <reiddraper> nahamu: can i use something with vmadm send and receive to clone the vm?[00:33:50] <nahamu> possibly. I haven't really played with that.[00:33:57] <nahamu> gotta run![00:34:41] <reiddraper> thanks[01:18:45] *** dis has quit IRC[01:20:35] *** dis has joined #smartos[01:40:23] *** ira has quit IRC[01:49:04] *** enmand has joined #smartos[01:58:07] *** enmand has quit IRC[02:01:24] *** MacSpengo has joined #smartos[02:02:31] <MacSpengo> Is there a good way to go about creating a Windows XP VM? I am having problems with video, network, and detecting a Hard Drive[02:02:53] <LeftWing> By problems do you mean it hangs during the install?[02:09:08] <MacSpengo> I dont even get that far. It tells me there is no drive detected, and through the VNC connection the screen is red with blue[02:10:09] <MacSpengo> how do I change the virtio hard drive I set to ide? Do I have to delete and recreate the vm with a corrected .json file?[02:10:34] *** jdavis has quit IRC[02:12:51] <LeftWing> I'm not sure, but I suspect you do.[02:13:33] <LeftWing> You will pretty much definitely need an IDE disk rather than virtio for XP, though, so that's a good start. I would make sure you're not using a virtio NIC at the same time.[02:16:38] <MacSpengo> I am, just noticed it. What should I use for the nic?[02:17:10] <MacSpengo> No way to update or change this configuration without delete the VM and recreating it from a .json file?[02:17:44] <rmustacc> You probably want to use e1000g[02:17:45] *** darjeeling has quit IRC[02:17:51] <rmustacc> Or e1000.[02:17:55] <rmustacc> Don't remember which its called there.[02:18:08] <rmustacc> And you can update it via vmadm update[02:22:20] <MacSpengo> rmustacc: I tried using the vmadm command. But it didnt apprear to change anything when i ran vmadm get[02:23:02] <MacSpengo> rmustacc: vmadm change UUID nics_model=e1000, but it still says virtio[02:25:46] <rmustacc> I assume you mean vmadm update. But you should probably look at the man page example, as I'm not sure that will be sufficient.[02:32:07] *** enmand has joined #smartos[02:50:20] *** dap has quit IRC[02:50:27] *** banjiewen is now known as banjiewen|away[03:05:21] <MacSpengo> yes, still cannot figure it out using vmadm update[03:06:45] *** darjeeling has joined #smartos[03:16:31] *** deirdres has quit IRC[03:16:46] *** sebasp has quit IRC[03:17:27] *** sebasp has joined #smartos[03:18:50] *** guk has quit IRC[03:20:17] *** guk has joined #smartos[03:20:31] *** bradleymeck_ has joined #smartos[03:22:38] *** bradleymeck has quit IRC[03:22:38] *** bradleymeck_ is now known as bradleymeck[03:27:05] *** deirdres has joined #smartos[03:29:43] *** jamesd__ has quit IRC[03:37:07] *** dis has quit IRC[03:39:16] *** dis has joined #smartos[03:46:39] *** sam_wan has joined #smartos[04:16:29] *** ryancnelson has joined #smartos[04:18:37] *** dis has quit IRC[04:20:26] *** dis has joined #smartos[04:30:07] *** dis has quit IRC[04:31:31] *** enmand has quit IRC[04:32:03] *** dis has joined #smartos[04:37:02] *** ryancnelson has quit IRC[04:54:03] *** dis has quit IRC[04:55:45] *** dis has joined #smartos[05:04:32] *** dis has quit IRC[05:05:51] *** ira has joined #smartos[05:06:25] *** dis has joined #smartos[05:14:30] *** bixu has joined #smartos[05:15:22] *** jdavis has joined #smartos[05:25:17] *** ira has quit IRC[05:31:18] *** dis has quit IRC[05:33:15] *** dis has joined #smartos[05:46:26] *** dis has quit IRC[05:46:54] *** denizr has quit IRC[05:47:15] *** denizr has joined #smartos[05:48:56] *** dis has joined #smartos[05:58:17] *** jdavis has quit IRC[06:18:50] *** denizr has quit IRC[07:09:56] *** doki_pen has quit IRC[07:18:47] *** doki_pen has joined #smartos[07:27:49] *** deirdres has quit IRC[07:38:00] *** sam_wan1 has joined #smartos[07:40:37] *** sam_wan has quit IRC[08:11:13] <MerlinDMC> morning[08:18:13] *** mamash has left #smartos[08:44:05] *** marsell has quit IRC[09:01:09] *** doki_pen has quit IRC[09:07:21] *** doki_pen has joined #smartos[09:10:27] *** doki_pen has quit IRC[09:18:43] *** doki_pen has joined #smartos[09:22:14] *** bens1 has joined #smartos[09:24:37] *** CnC has joined #smartos[09:25:13] *** mamash has joined #smartos[09:36:47] *** doki_pen has quit IRC[09:37:40] *** doki_pen has joined #smartos[09:40:31] *** CnC has quit IRC[10:24:04] *** marsell has joined #smartos[10:45:33] *** akole has quit IRC[10:46:27] *** akole has joined #smartos[11:02:21] *** Licenser has joined #smartos[11:02:36] *** sam_wan1 has quit IRC[11:06:13] <Licenser> goood morning #SmartOS![11:10:07] *** nicksellen_ has joined #smartos[11:10:52] *** nicksellen_ has left #smartos[11:15:09] *** nicksellen has joined #smartos[11:17:00] *** max has joined #smartos[11:17:23] <max> hi[11:17:24] *** max is now known as Guest18491[11:17:29] *** Guest18491 is now known as max`[11:18:21] <Licenser> hi max :)[11:20:13] *** AikoM has joined #smartos[11:20:15] <trentster> MerlinDMC: howdy you around?[11:25:18] *** mamash has left #smartos[11:32:28] <MerlinDMC> what?[11:32:41] <MerlinDMC> i'm at work ... so i have to be here somewhere ;9[11:33:46] *** darjeeling has quit IRC[11:35:06] <MerlinDMC> trentster, that "howdy" makes me think about bbq ... and i'm not able to the next 8 hours -.- now i'm kinda hungry[11:35:37] <Licenser> MerlinDMC it's called brie for him not BBQ :P[11:35:56] <Licenser> we'll have brie on frieday ro suttarday again I think[11:36:06] <Licenser> good reason to stay at least untill start of next week :P[11:38:24] *** saablover has joined #smartos[11:39:41] <saablover> hi people, what does the line dns_resolvers= mean in the /usbkey/config ?[11:41:35] <Licenser> saablover I think it's like DNS servers :)[11:42:26] <saablover> yeah correct, but when I leave it as is I can reach my dns server, or is this only needed for the zones ?[11:43:41] *** mamash has joined #smartos[11:43:57] <Licenser> what do you mean with 'leave it as it is' and 'reach your dns server'[11:44:04] * Licenser is a bit confused now[11:44:46] <saablover> just a second[11:47:05] *** mamash has left #smartos[11:47:05] <saablover> at the moment the machine has this configuration: /usbkey/config dns_resolvers=8.8.8.8,8.8.4.4[11:47:18] <saablover> I don't have any zones installed[11:47:39] <saablover> ping www.google.com[11:47:39] <saablover> www.google.com is alive[11:47:42] <saablover> this works[11:47:59] <saablover> so I assume that this line influences the zones configuration[11:49:39] *** mamash has joined #smartos[11:54:03] <MerlinDMC> saablover, these are the dns server to use for the GZ[11:54:31] <MerlinDMC> those are written to /etc/resolv.conf at boot[11:55:08] <saablover> my admin interface is currently set to dhcp and /etc/resolv.conf is filled in by dhcp at the moment[11:55:13] <MerlinDMC> (only if you have an admin nictag - but that is created by default)[11:55:16] <saablover> so this is for a static configuration ?[11:55:52] <MerlinDMC> this would be used for static configuration yes[11:55:58] <saablover> ok that's clear[11:56:08] <saablover> when I configure a zone, it picks up the same config ?[11:56:28] <MerlinDMC> you set the resolvers for a zone in the json file[11:56:59] <saablover> ok clear[11:57:03] <saablover> thx for the help[11:57:19] <MerlinDMC> no problem[11:57:49] <MerlinDMC> Licenser, got it now ... or totally confused? :)[11:57:52] <trentster> Hi MerlinDMC, Licenser mentioned you use freebsd on smartos, is that correct?[11:58:14] <MerlinDMC> trentster, yep fbsd9 with virtio drivers[11:58:53] <Licenser> MerlinDMC yea I was just wondering about saablover's setup :)_ and then got called away for work :([11:59:00] <trentster> I have been testing freebsd 9 with virtio as well,[11:59:47] <trentster> both with virtio packages and with compiling it one the machine. In both cases the disk io benchmarks are still very very slow compared to linux virtio on smartos, do you find this as well?[12:00:14] <MerlinDMC> trentster, i don't benchmark as long as it is "fast enough for me"[12:00:31] *** saablover has quit IRC[12:00:42] <MerlinDMC> and my fbsd box does more or less noting else than serving teamspeak and smtp[12:01:16] <MerlinDMC> but if you send me what you did for benchmark i could test it on my box to compare values[12:01:16] <trentster> ok, got it.[12:01:47] <MerlinDMC> i have also some other box with fbsd on linux host i think ... let me see if that uses virtio[12:02:04] <trentster> just did a dd on local disk[12:02:27] *** saablover has joined #smartos[12:02:41] <MerlinDMC> trentster, what block size?[12:03:43] <trentster> 8k and 128k[12:05:16] <max`> killfill: You around?[12:07:39] *** saablover has quit IRC[12:09:11] *** saablover has joined #smartos[12:14:43] *** saablover has quit IRC[12:25:50] <Licenser> dun dun dun :)[13:01:23] *** dis has quit IRC[13:08:44] *** dis has joined #smartos[13:19:08] *** enmand has joined #smartos[13:44:07] *** sebash has joined #smartos[13:44:11] *** sebash is now known as fooba_[13:46:52] *** mamash has left #smartos[13:47:56] *** fooba_ is now known as foob[13:48:01] <foob> .fooob[13:48:10] *** foob is now known as fooob[13:53:03] *** max` has quit IRC[14:09:11] <jeffpc> the admin interface should be on the LAN, right?[14:09:17] <jeffpc> (not internet facing)[14:12:08] <MerlinDMC> jeffpc, private network :)[14:12:16] <jeffpc> yeah, thought so[14:16:40] <fooob> I am still trying to install smartos completely remote, is there a way to start a SSH server and DHCP using the noinstall boot option? Maybe by altering the image?[14:17:06] <fooob> then I could just start the setup script via ssh[14:19:35] <jeffpc> MerlinDMC: so, in Joyent's case, the admin network is for joyent staff, and the "other" network is for customers[14:20:04] <jeffpc> since I want to do NAT on the box too, I'll end up with 3 networks in the GZ: admin, internal, external[14:20:09] <jeffpc> correct?[14:23:17] <MerlinDMC> jeffpc, admin + external would be enough ... but it's up to you how many networks you populate[14:23:54] <MerlinDMC> on some of my boxes i'm limited to one nic ... so i've admin + external on one physical network and an etherstub for nat[14:24:34] *** bradleymeck has quit IRC[14:24:53] <jeffpc> I have two ports, so I'm thinking about using one instead of an etherstub...which will also let me plug in my laptop when I'm in the datacenter[14:25:52] <jeffpc> is "admin" essentially "global zone network"?[14:25:52] *** fooob has quit IRC[14:26:32] <jeffpc> (I have only one box)[14:26:44] <MerlinDMC> most of the prebuild datasets require two interfaces riak for example ... and for riak clustering the admin network is used ...[14:27:25] <MerlinDMC> in most cases it's just the network where critical data can be send from a to b without extra care i guess[14:27:52] <MerlinDMC> but maybe ryan can tell you more on that (i hope it was ryan)[14:29:01] <MerlinDMC> this one: http://www.youtube.com/watch?v=TIdbggrEZMY ... and it was ryan :)[14:30:35] * jeffpc watches[14:35:41] <jeffpc> MerlinDMC: it's a nice video![14:52:04] *** jamesd__ has joined #smartos[15:06:30] *** mamash has joined #smartos[15:06:52] <nahamu> mamash: I put the content from your email into the wiki as a place holder.[15:07:53] <mamash> thanks, nahamu[15:12:47] <nahamu> also, hi. :)[15:18:02] *** bradleymeck has joined #smartos[15:25:18] *** echelog-1 has joined #smartos[15:32:13] *** lpin has joined #smartos[15:33:39] <jeffpc> Licenser: fifofy (dev) doesn't quite like if you accidentally use an in-use IP address[15:33:50] <jeffpc> Licenser: it barfs on vm creation, but then tries to go on[15:34:06] <Licenser> jeffpc thanks I really should catch that :) you're right[15:34:12] *** bradleymeck has quit IRC[15:34:24] <jeffpc> Licenser: I'm trying it on real hardware :)[15:34:34] <Licenser> heh[15:34:35] <jeffpc> cp: cannot access /zones/fifo/root/fifo/sniffle.tar.bz2[15:34:37] <Licenser> cool cool :)[15:34:47] <Licenser> you've propably net set your gw/dns correctly[15:35:08] <jeffpc> gw & dns are set right[15:35:23] <jeffpc> it's installing other things just fine[15:35:29] <jeffpc> (python, etc.)[15:36:10] <Licenser> oh wait it failed with the zone creation right?[15:36:42] <jeffpc> oh, sorry... the zone creating failed when I gave it the wrong (in-use) IP for the ui zone[15:36:52] <jeffpc> I cleaned it up, and it got further[15:37:29] <jeffpc> http://pastie.org/4237789[15:37:48] <jeffpc> http://pastie.org/4237794[15:38:23] <Licenser> give me like 30 minutes I'm in the middle of something is that OK?[15:46:08] <jeffpc> I'll have to leave for work soon[15:46:59] <Licenser> can you mail me the file /var/log/fifo-install.log ?[15:47:06] <jeffpc> sure[15:47:11] <Licenser> or just do ./fofofy.sh collect[15:47:18] <Licenser> you get a neatly packed archive with all[15:47:24] <Licenser> just mail it to heinz at licenser dot net :)[15:47:34] <Licenser> and I'll mail you with what went wrong :)[15:47:49] * Licenser has to head out of the center now -.- customer is kicking me out[15:51:45] *** Licenser has quit IRC[15:54:26] *** bens1 has quit IRC[15:55:10] <noahmehl> I'm having trouble using the "image_uuid"[15:55:11] <noahmehl> http://pastebin.com/e69nYekG[15:56:29] <MerlinDMC> noahmehl, discard the "-disk0" i guess[15:57:41] *** cypres has quit IRC[16:02:44] <noahmehl> that allowed me to make the machine[16:02:50] <noahmehl> but now it doesn't boot :([16:04:18] <noahmehl> and now the uuid-disk0 is a filesystem and not a device[16:04:22] <nahamu> noahmehl: are you trying to clone an existing one?[16:04:27] <noahmehl> yes[16:04:41] <noahmehl> like, machine A: has a kvm disk[16:04:44] <nahamu> the wiki page has notes on what has changed.[16:04:52] <noahmehl> which wiki page?[16:04:57] <nahamu> one sec...[16:05:19] <noahmehl> http://wiki.smartos.org/display/DOC/Managing+Datasets[16:05:20] <noahmehl> ?[16:05:27] <nahamu> probably[16:06:27] *** CarlosC has joined #smartos[16:08:16] <nahamu> basically you need to clone your old <uuid1>-disk0 into a <uuid2> and use that second uuid as the basis of your other VMs[16:08:42] <nahamu> I'm the one who fixed up that web page so if it isn't clear let me know.[16:13:21] *** Licenser has joined #smartos[16:13:42] <Licenser> there we go I'm back[16:15:54] <nahamu> noahmehl: did you get it working?[16:16:44] <noahmehl> nahamu: still working[16:16:46] <noahmehl> but I think i get it[16:16:50] <noahmehl> i'm cloning a uuid-disk0[16:16:52] <noahmehl> to a uuid[16:16:59] <noahmehl> then creating a vm with that image_uuid[16:17:13] <noahmehl> so that the resultant uuid-disk0 is a spare image and not a filesystem?[16:18:01] <noahmehl> nahamu: looking better![16:18:48] <noahmehl> yes, that seemed to work[16:18:50] <noahmehl> YAY![16:18:51] <noahmehl> thanks![16:23:19] *** marsell has quit IRC[16:24:51] *** PiotrSikora has quit IRC[16:25:05] *** PiotrSikora has joined #smartos[16:26:20] <nahamu> noahmehl: no problem. :)[16:31:25] *** ira has joined #smartos[16:34:12] *** denizr has joined #smartos[16:36:02] <Licenser> jeffpc still around?[16:37:21] <jeffpc> Licenser: I'm already at work; so yes, but without access to the test box[16:37:21] *** jamesd_ has joined #smartos[16:37:41] <Licenser> ah okay :) I'll push a new install script that exits on errors[16:40:07] *** jamesd__ has quit IRC[16:40:20] <jeffpc> Licenser: how does one uninstall?[16:40:31] <Licenser> ./fifofy.sh uninstall[16:40:46] <jeffpc> oh, nice[16:41:24] <Licenser> jeffpc it's pure egoism :P as much as in install/uninstall fifo I HAVE to have a script for that[16:42:20] <Licenser> [ERROR] Failed to create fifo zone! :D[16:44:13] *** denizr has quit IRC[16:44:42] *** denizr has joined #smartos[16:44:59] *** bens1 has joined #smartos[16:50:23] *** denizr1 has joined #smartos[16:50:26] *** lpin_ has joined #smartos[16:50:29] *** lpin has quit IRC[16:54:13] *** denizr has quit IRC[16:59:34] *** darjeeling has joined #smartos[17:01:32] *** jdavis has joined #smartos[17:01:40] *** lpin_ is now known as lpin[17:13:39] <killfill> hm.. i just imported centos6 sdc:jpc:centos-6:1.0.1.. it it listed in datasets list, but when creating a VM on SDC GUI, i see just ubuntu..[17:22:04] *** deirdres has joined #smartos[17:25:20] *** mamash has left #smartos[17:25:23] *** bens1_ has joined #smartos[17:27:17] *** bens1 has quit IRC[17:29:11] *** bixu has quit IRC[17:30:56] *** bens1_ is now known as bens1[17:31:23] *** deirdres has quit IRC[17:35:13] *** max has joined #smartos[17:35:36] *** max is now known as Guest49593[17:35:38] *** Guest49593 is now known as max`[17:36:43] *** MacSpengo has quit IRC[17:39:24] *** AikoM has quit IRC[17:48:01] <Triskelios> is there a procedure to update an existing zone to a newer image?[17:48:53] *** Licenser has quit IRC[17:50:04] *** sjorge has joined #smartos[17:51:49] <Triskelios> nevermind, I think should just pull an image into a dataset and clone it[17:55:30] <rmustacc> Triskelios: a joyent branded zone or a kvm one?[17:55:53] <Triskelios> a Joyent-branded zone[17:56:20] <rmustacc> There are scripts that can do the update from one pkgsrc repo to another, assuming that's what you want.[17:57:34] <Triskelios> I want to update the core userland and there's no data in the zone so I can just discard the current root[17:57:59] <rmustacc> As in, you don't care about pkgsrc, you just want to update the sparse pieces?[17:58:13] <rmustacc> If so, that's just a reboot of the host.[17:59:26] <Triskelios> er? they're not sparse zones and the host has been rebooted already[18:00:07] <e^ipi> they are sparse zones.[18:00:18] <Triskelios> planning to do: imgadm import <new image>, zfs destroy <old zone dataset>; zfs snapshot <new image dataset>; zfs clone <snapshot of new image dataset> <zone dataset name>[18:00:41] *** dap has joined #smartos[18:01:53] <Triskelios> e^ipi: oh, wow. guess I can just sync the out of date files in /etc from the host then[18:01:57] <Triskelios> thanks[18:02:16] <rmustacc> Files in /etc are their own thing.[18:03:48] <Triskelios> gotcha[18:04:15] <rmustacc> Because /etc can't be sparse.[18:27:30] *** deirdres has joined #smartos[18:28:53] *** max` has quit IRC[18:29:05] *** mamash has joined #smartos[18:37:38] *** Saskaloon has quit IRC[18:37:57] <nahamu> mamash: Does (or will) sm-prepare-image recreate the /root/zoneinit* bits and the zoneinit service?[18:40:01] *** Saskaloon has joined #smartos[18:43:59] *** dap has quit IRC[19:09:45] <mamash> nahamu: yes, under /var/zoneinit though[19:10:14] <mamash> if you don't let it shut down the zone at the end, you can go into /var/zoneinit yourself and add stuff to happen at provision time[19:10:22] *** denizr1 has quit IRC[19:10:44] *** denizr has joined #smartos[19:27:36] <CarlosC> does anyone have a build based on base-1.7.0?[19:27:53] *** Licenser has joined #smartos[19:28:56] <CarlosC> or better yet, with this as your pkg source: 2012Q1[19:29:10] <CarlosC> looks like mpfr is missing and is required by build[19:29:20] <wesolows> a build of what?[19:29:31] <nahamu> mamash: cool, thanks![19:29:51] <CarlosC> wesolows smartos[19:30:24] <wesolows> I haven't tried to build on 1.7 yet, no. But I thought we installed mpfr for the gcc build.[19:31:03] <CarlosC> I thought so as well…do you know where it might have been installed?[19:32:46] <rmustacc> Some things changed in 1.7, so there could very well be changes necessary.[19:34:37] <Licenser> dun dun dun :)[19:35:53] <rmustacc> Specifically there was changes with respect to where gcc put some of its files, so who knows.[19:36:12] <CarlosC> yeah…going to start "combing the desert"[19:36:41] <CarlosC> if all else fails, I'll spin up another build zone based on another dataset[19:36:59] <rmustacc> Unless you really want to blaze that trail, I would stick to 1.6.[19:37:46] <CarlosC> my afternoon if open ;)[19:37:56] <Licenser> CarlosC can you come by and clean my place?[19:38:05] <CarlosC> hah ;)[19:47:04] *** dap has joined #smartos[19:48:59] <mamash> CarlosC: mpfr is only required for GCC in pkgsrc, and we prefer to have GCC build it embedded[19:49:08] <mamash> i can add the package separately though[19:49:47] <CarlosC> so, does the gcc in illumos-extra need to be updated to reflect this?[19:51:40] *** denizr has quit IRC[19:52:06] *** denizr has joined #smartos[19:55:18] <rmustacc> mamash: Can you package it? I think we rely on it for building the image.[19:55:25] <rmustacc> mamash: Though I need to double check with wesolows.[19:59:38] <wesolows> the gcc in illumos-extra is set up to use mpfr from pkgsrc, which is correct because it runs on the build system.[20:01:39] <jperkin> just wish gcc wasn't such a pig these days, takes forever to build[20:01:41] <wesolows> since there's no other source for mpfr and gcc 4.4.4 requires it to build, it needs to be in pkgsrc. And I just verified that configure will indeed try to install it.[20:01:52] <wesolows> it takes about 10 minutes for me :-)[20:02:25] <wesolows> it's gotten bigger much, much more slowly than machines have gotten faster at building... and let's not forget that in the egcs era it also couldn't build with make -j.[20:02:45] * wesolows remembers building egcs on an Indy r5k[20:04:29] <CarlosC> wesolows what's the long term goal for compiler support in regards to building smartos: stick with gcc or move to clang?[20:04:42] <wesolows> there's no plan to move to clang.[20:04:46] <CarlosC> k[20:05:10] <jperkin> however, should that change, pkgsrc is actively being worked on for clang support :)[20:05:11] <wesolows> gcc is working fine. we'd happily support clang if it worked well, especially if it could generate lint libraries, but it's not something we're investing in[20:05:42] <wesolows> (and yeah, in case the distinction isn't clear, jperkin is talking about pkgsrc and I'm talking about the platform -- they can and may have different stories there)[20:05:44] <jperkin> mmm egcs[20:06:12] <CarlosC> np…just curios after I read the Freebsd post[20:07:08] <CarlosC> http://www.phoronix.com/scan.php?page=news_item&px=MTEwMjI[20:09:19] <wesolows> personally I have no desire to work on a compiler or anything else written in C++[20:09:30] <CarlosC> ;)[20:09:31] <rmustacc> CarlosC: A lot of work and modifications have to be made to clang to get it to build illumos. Further, it doesn't work on illumos wright now, or at least, not well is my impression.[20:09:34] <wesolows> gcc would have to fail pretty horribly for me to think about clang[20:28:26] *** Licenser has quit IRC[20:29:39] <e^ipi> wesolows: but i thought endless churn based on what the latest unproven hipster hotness was what this industry was about[20:30:49] *** bens1 has quit IRC[20:30:50] *** Licenser has joined #smartos[20:31:07] <CarlosC> e^ipi maybe if it's web based ;)[20:38:50] *** ira has quit IRC[20:45:51] *** Niamkik has joined #smartos[20:58:54] *** Licenser has quit IRC[20:59:47] <wesolows> e^ipi: you're right of course. That's why I'm rewriting SmartOS in clojerlscriptrubylangmotion[21:00:08] <wesolows> I heard some people talking about it at Four Barrel[21:01:35] <bdha> ERLANG ALL THE THINGS[21:04:17] *** Licenser has joined #smartos[21:04:41] *** Niamkik has quit IRC[21:06:18] *** deirdres has quit IRC[21:11:58] <killfill> can it be that a ubuntu VM is quickier than a SmartMashine?[21:14:16] <nahamu> killfill: for certain things, potentially. what are you seeing?[21:16:31] <killfill> nahamu: im playing with sysbench.. a simple mysql test, shows its faster when using more that 20 thread.[21:16:52] <killfill> nahamu, and a cpu test shows its faster too.[21:16:53] <killfill> like 20%?[21:17:09] <killfill> im creating the mashines with the same package[21:17:31] <killfill> smartos is cpu capped to 400, and vcpu is 4[21:18:26] <killfill> nahamu: i thought smartos vm's would be alwais quickier, becouse of the virtualization?[21:22:57] <wesolows> we'd really have to dig into it to know what you're seeing[21:23:10] <wesolows> I think Brendan is looking at a few things that might be similar.[21:24:34] <nahamu> I vaguely remember reading somewhere that Linux syscalls are faster than Solaris ones which may or may not have any shred of truth to it... But if so, you'd expect that a largely CPU bound benchmark that hits lots of syscalls might indeed do better in an ubuntu VM under KVM...[21:25:20] <nahamu> but until Bryan figures out how to dtrace a guest OS from the GZ, if you're ever trying to figure out *why* things are slow... I'd rather take a small performance hit and gain the visibility.[21:26:19] <killfill> hm.. ok i notices the ubuntu image was 64bit.. i repeted the test with a smartos64, and the results are equal..[21:26:43] <nahamu> should have kept my mouth shut... :)[21:26:53] <killfill> hehe.. :P[21:27:00] <killfill> well but the mysql test are much slower[21:27:42] <killfill> i.e. with 50 threads on linux i get 798tps, and on smartos 478tps[21:30:51] * killfill wonders how to check if your on a 32 or 64 bit smartos[21:31:13] <nahamu> the kernel is always 64 bit and can run 32bit applications[21:31:42] <nahamu> so it's whether the binaries in /opt/local/bin are 32 or 64 bit[21:31:43] <killfill> but i.e. you have 'base' and 'base64'.. wondering what is the 'mysql' dataset[21:31:52] <nahamu> and I think you could just run 'file' on them and find out.[21:32:21] <nahamu> good question[21:32:50] <killfill> hm.. on base64, /bin/ls is 32bit ELF[21:33:19] <nahamu> you need to look in /opt/local[21:33:34] <nahamu> /bin comes from the platform[21:35:44] <killfill> Ah, i guess its the becouse its the lowest common denominator[21:35:57] <killfill> anyway.. mysql is 32 bit.. :P[21:36:30] <jperkin> not quite, utils in the platform which can make use of 64bit are built multiarch, and the executable is a link to isaexec which will effectively always pick 64-bit on SmartOS[21:36:56] <killfill> ah[21:37:41] <jperkin> pkgsrc does not yet have support for multiarch, hence us providing datasets for 32-bit or 64-bit, but we're working to fix that.[21:38:37] <killfill> aah, thats why a see a different version of a package on 32 and 64 bases[21:39:43] <killfill> (package i mean pksrc package..:P)[21:45:34] <noahmehl> nahamu: I'm on my second machine from the cloned source[21:45:42] <noahmehl> but I can't seem to connect to the vnc[21:45:47] <noahmehl> where would the log for the machine be?[21:45:48] <konobi> blocksize might make a difference too[21:46:02] <killfill> hm.. the percona dataset has actually a 64bit percona binary...[21:46:19] <killfill> dont know why the linux mysql outperforms percona.. :S[21:46:28] <nahamu> noahmehl: /zones/<uuid>/root/var/tmp/vm.log (I think)[21:46:36] <nahamu> err, hang on[21:46:44] <killfill> konobi: blocksize?[21:47:10] <konobi> filesystem blocksize[21:47:16] <nahamu> no var in there[21:47:25] <noahmehl> i found it[21:47:27] <noahmehl> thanks![21:47:33] <nahamu> /zones/<uuid>/root/tmp/vm.log[21:47:35] <nahamu> good[21:48:34] <killfill> hmm...[21:49:48] <killfill> i wonder if on joyent's public cloud, ubuntu+mysql works better than a percona image, like in my setup[21:52:29] <nahamu> if anyone would know, it's brendang... :-P[21:58:34] <killfill> i remember mysql was alwais slower on freebsd than linux becouse it uses a time syscall[21:58:46] <killfill> wich was more slow (an exact) on freebsd[21:58:57] <killfill> maybe its the case for smartos too..[22:01:06] *** bens1 has joined #smartos[22:03:30] *** bens1 has quit IRC[22:05:18] *** Licenser has quit IRC[22:09:16] *** Licenser has joined #smartos[22:15:46] *** chorrell has joined #smartos[22:17:24] <killfill> http://jsfiddle.net/t8EtG/ <-- that my humble results.. :P[22:17:56] <killfill> <20 threads, ubuntu is winning.. :S[22:18:37] *** enmand has quit IRC[22:19:23] <nahamu> but if percona with 10 threads performs better than ubuntu with 20...[22:19:49] <rmustacc> Are you sure you have the actual same configuration on the database?[22:20:02] <killfill> yup, same config[22:20:06] <rmustacc> You aren't really doing an apples to apples test, which makes it rather hard to actually understand what the problem is.[22:20:17] <rmustacc> You're using percona on boht?[22:20:18] <rmustacc> *both[22:20:33] <killfill> yah i know.. thats why it a humble test.. :P[22:20:43] <killfill> rmustacc: on both?[22:20:53] <rmustacc> On both Ubuntu and in your SmartMachine.[22:21:07] <killfill> ah no, ubuntu is using mysql[22:21:14] <rmustacc> Then you're not testing the same thing.[22:21:36] <rmustacc> If you're asking why performence is different and actually want an actual answer, you need to be really rigorous.[22:21:36] <killfill> ya, thats why i test mysql@sdc too[22:21:57] <rmustacc> And you have to figure out why they're different and not why they're different.[22:22:18] *** knweiss has joined #smartos[22:22:31] <rmustacc> For example, when Brendan and I were testing kvm, we found that our KVM implementation caused a CPU bound workload to be baster than Linux bare metal. Which of course, if you think about is completely wrong.[22:22:55] <rmustacc> So benchmarking is really all about understanding why it's different.[22:23:30] <killfill> yeah, ill try to make it better.[22:23:40] <rmustacc> In that case it turned out one machine had Intel turbo boost turned on, the other didn't, and it was in a cooler part of the data center.[22:23:55] <killfill> oh :)[22:24:09] <rmustacc> So sometimes the answers are really weir.d[22:24:11] <rmustacc> *weird[22:24:29] <rmustacc> Once it was someone built packages with gcc 3.x and someone with gcc 4.x. Bit of a difference there.[22:25:19] <nahamu> on the bright side, if you're just looking for which of your configuration options was fastest, it's nice to see that percona with 10 threads on SDC came out on top. :)[22:26:00] <killfill> heh..[22:26:09] <rmustacc> That's some interesting data.[22:26:14] <nahamu> is the number of threads the number of client threads?[22:26:15] <rmustacc> How many CPUs do you have on that box again?[22:26:42] <nahamu> or the number of server threads?[22:27:11] <killfill> im actually 'fighing' for the xen there is on the same dell310... wanted to cerify thing were working better (wich they are..) but the linux on sdc test result was not expected...[22:27:31] <killfill> (i mean no on the same.. a clone)...[22:27:56] <rmustacc> Seems like a case of scheduling differences, would be my guess there.[22:28:04] <killfill> rmustacc: its a 1 xeon, 8 'cpu's[22:28:15] <rmustacc> That's a useful data point though for us to keep in the back of our heads, thanks.[22:28:19] <killfill> capu-cap and vcpu for the vm's are 400 and 4.[22:28:25] <nahamu> interesting that both Xen VMs came out identical.[22:28:47] <killfill> hm..[22:28:59] <killfill> of sorry.. http://jsfiddle.net/t8EtG/2/ only the first is xen...[22:29:09] <killfill> al others is on sdc.. sorry about the type[22:29:31] <killfill> nahamu, is the number of threads the number of client threads? --> the client threads[22:30:31] <killfill> ill better re-test with more detail, and send en email to the list.. :)[22:31:00] <nahamu> ah[22:37:04] *** noahmehl has quit IRC[22:49:49] *** nicksellen has quit IRC[23:00:12] *** bens1 has joined #smartos[23:00:40] *** Licenser has quit IRC[23:06:42] *** Licenser has joined #smartos[23:06:58] *** bens1 has quit IRC[23:16:08] *** dap has quit IRC[23:16:51] *** dap has joined #smartos[23:23:34] *** denizr has quit IRC[23:24:06] *** denizr has joined #smartos[23:33:29] *** Meths has quit IRC[23:34:39] *** Meths has joined #smartos[23:35:34] *** Meths has quit IRC[23:43:45] *** Meths has joined #smartos[23:43:53] *** Licenser has quit IRC[23:44:34] *** Licenser has joined #smartos[23:51:31] *** CnC has joined #smartos[23:59:27] *** CarlosC has quit IRC