July 26, 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 26, 2012 [00:05:57] *** ira has quit IRC[00:05:57] *** ira_ is now known as ira[01:00:33] *** wdent has joined #smartos[02:11:47] *** denizr has quit IRC[02:55:06] <trentster1> rmustacc: thanks for the "nicstat"[02:58:02] <rmustacc> Yeah. Should probalby integrate that or figure out what's up with dlstat.[02:58:50] <richlowe> what's the symptom?[03:00:16] <rmustacc> https://github.com/joyent/smartos-live/issues/102[03:01:11] <richlowe> if I were taking wild guesses, they'd involve having so many vnics with the same name[03:01:16] <richlowe> I do mean _wild_ guesses though[03:01:17] <rmustacc> Basically things don't show up.[03:01:24] <rmustacc> Well, we are a big vanity naming shop.[03:01:46] <rmustacc> So that could do it.[03:01:49] <richlowe> I thought allowing multiple instances of a name was something you guys added, too[03:01:58] <richlowe> I recall being told solaris does it, but differently, at least.[03:01:59] <rmustacc> Yeah[03:02:05] <trentster1> how do I correlate an int name e.g. " z19_net0 " with a zone uuid ?[03:02:38] <rmustacc> Run zoneadm list -v[03:03:10] <rmustacc> But the zonename should also be in some of the kstats.[03:03:15] <trentster1> aaah thanks[03:03:17] <rmustacc> As an actual field.[03:03:29] <rmustacc> Note that those zone ids do change based on zone reboots.[03:04:11] <trentster1> whats a e.g. syntax using kstat to pull just zone nic stats and display it sanely?[03:05:34] <rmustacc> Sorry, I'm not the one to help there.[03:05:40] <trentster1> or should I say what is a reliable way to get accurate vnic stats that will survive zone reboots (ID changes) that can be used for ip accounting/ reporting?[03:05:51] <rmustacc> richlowe: I do wnat to dig into it at some point, but higher priorities and all those fun things.[03:06:34] <richlowe> Yeah, I just wondered what was happening[03:06:43] <richlowe> I want to pull your libdl* changes, basically[03:06:47] <richlowe> I know Jerry fixed a lot in there[03:06:54] <richlowe> so it's more worrying that this will stop me :)[03:07:46] <rmustacc> Some day I really need to go through and just start categorizing all the unupstreamed stuff.[03:08:32] <rmustacc> trentster1: Well, are you doing it from inside the zone or from the GZ?[03:08:34] <richlowe> John was going to do something like that, but he got bored of us again I think[03:08:45] <trentster1> rmustacc: from the GZ[03:09:14] <rmustacc> Honestly, probably using libkstat in some form, like node-kstat, and grabbing out the zonename field.[03:09:21] <rmustacc> And aggregating using that.[03:09:43] <trentster1> ok, thanks[03:10:40] <rmustacc> There are probably other ways.[03:11:08] <rmustacc> Maybe netstat has something that you're looking for.[03:11:15] <rmustacc> I don't know if that's zone aware though.[03:11:47] <rmustacc> If you're looking at nicstat you could probably modify something like that to change the display name it's using based on a zonename field.[03:13:33] *** dap has quit IRC[03:14:14] <rmustacc> richlowe: As a small survey, when's the last time you had PCI-X cards (I do not mean PCIe).[03:16:25] <rmustacc> I'm trying to convince myself I should care about a device using not MSI-Xs.[03:17:13] <richlowe> I don't recall[03:17:45] *** dap has joined #smartos[03:18:08] <rmustacc> And really, the laziness comes from not wanting to test everything in msi/intx mode.[03:18:17] *** dap has quit IRC[03:18:52] <richlowe> it may help to know what you're changing.[03:19:34] *** relling has quit IRC[03:29:15] *** denizr has joined #smartos[03:35:39] *** denizr has quit IRC[03:43:58] *** denizr has joined #smartos[04:03:12] *** relling has joined #smartos[04:09:45] *** jdavis1 has quit IRC[04:16:42] *** denizr has quit IRC[04:18:32] *** ira has quit IRC[04:27:30] *** kfr has quit IRC[04:38:59] *** Webhostbudd has joined #smartos[04:49:24] *** enmand has quit IRC[05:20:30] *** denizr has joined #smartos[05:25:49] *** denizr has quit IRC[05:27:38] *** elijah-mbp has quit IRC[05:27:57] <siezer> hello. I'm trying to "slipstream" a USB driver onto a USB image.... is there anything more I should have to do than this? http://bit.ly/NW2SsG[05:28:10] <siezer> (please excuse my deranged shell scripting style)[05:30:50] <siezer> er, USB nic[05:32:34] <siezer> the driver works when I install it in a running smartos instance, but when I try and boot the image resulting from the above maneuver, the boot asplodes at the network services[05:33:37] *** aroop has joined #smartos[05:41:57] <Triskelios> picking a major number is kinda risky... it's more work, but you may want to futureproof this by just setting up an environment to build a fresh image[05:42:47] <siezer> yea.. unfortunately I'm trying to do this to an SDC image... no source =([05:45:10] <siezer> why is picking a major number risky?[05:50:11] <Triskelios> they're not fixed[05:55:44] <siezer> hrmmm.. well I'm just trying to duplicate what add_drv / update_drv was spitting out on a running machine (same results every time)[05:57:10] *** Sachiru has joined #smartos[05:57:44] <siezer> can you point me at a doc describing the boot process around device creation? oracle broke all the links on sun.com that I keep googling for[06:22:25] *** jelmd has quit IRC[06:33:29] *** jelmd has joined #smartos[06:39:07] *** pjama has quit IRC[06:51:06] *** Sachiru has quit IRC[06:53:44] *** pjama has joined #smartos[06:56:35] *** jamesd has joined #smartos[06:56:39] *** jelmd has quit IRC[06:59:41] *** jamesd_ has joined #smartos[07:02:51] *** jamesd has quit IRC[07:05:33] *** aroop_ has joined #smartos[07:05:33] *** aroop has quit IRC[07:10:10] *** marsell_ has joined #smartos[07:12:42] *** bixu has joined #smartos[07:13:26] *** marsell has quit IRC[07:13:27] *** marsell_ is now known as marsell[07:39:51] *** aroop has joined #smartos[07:41:09] *** aroop_ has quit IRC[07:59:44] *** elijah-mbp has joined #smartos[08:19:11] *** marsell has quit IRC[08:29:01] *** bixu has quit IRC[08:29:55] *** mathie has joined #smartos[08:34:16] *** mathie has quit IRC[08:43:26] *** bens1 has joined #smartos[09:08:53] <siezer> it totally worked btw[09:15:07] <MerlinDMC> that driver adding thing?[09:33:47] *** Tekni has quit IRC[09:34:42] *** mamash has joined #smartos[09:36:00] *** rodgort has quit IRC[09:36:52] *** rodgort has joined #smartos[09:40:50] *** konobi has quit IRC[09:41:00] *** konobi has joined #smartos[09:41:59] *** Tekni has joined #smartos[09:49:06] *** Gurpartap has quit IRC[10:16:49] *** Gurpartap has joined #smartos[10:20:30] *** marsell has joined #smartos[10:30:50] *** wdent has quit IRC[11:02:11] *** Z` has quit IRC[11:11:09] *** Licenser has joined #smartos[11:14:17] *** Z` has joined #smartos[11:17:08] <MerlinDMC> about 90 minutes to write a smartos usbkey ... that piece of hardware is horrible fast -.-[11:22:33] *** Z`` has joined #smartos[11:23:07] *** Licenser has quit IRC[11:26:03] *** Z` has quit IRC[11:27:43] <konobi> MerlinDMC: get patriot rage xt and use a more reasonable blocksize when dd-ing[11:27:55] <konobi> one of the conv options helps too[11:34:33] *** mamash has left #smartos[11:35:41] <MerlinDMC> konobi, i already have a blocksize of 1m ... on the other usbkeys that takes only some minutes to complete ...[11:36:06] <MerlinDMC> but yeah ... that one is for the office ... that can be slow as hell[11:36:33] <konobi> a usb HD might work too[11:37:48] <MerlinDMC> yep ... but that would look odd if i use tape to add that HD to the server front ;)[11:38:18] *** mamash has joined #smartos[11:45:05] <konobi> there's always the internal usb headers[11:58:35] *** aroop has quit IRC[12:06:46] *** Licenser has joined #smartos[12:06:53] <Licenser> yay[12:07:03] * Licenser reinstalled his system it's now up and working (properly agian!)[12:19:15] <MerlinDMC> with typo generator included![12:20:24] <trentster1> lol[12:28:56] *** mamash has left #smartos[12:33:32] *** mamash has joined #smartos[13:13:50] *** mamash has left #smartos[13:31:36] *** marsell_ has joined #smartos[13:34:57] *** marsell has quit IRC[13:34:58] *** marsell_ is now known as marsell[14:03:53] <Licenser> MerlinDMC totally[14:08:56] *** jrbing has quit IRC[14:09:26] *** jrbing has joined #smartos[15:12:08] *** jelmd has joined #smartos[15:27:43] *** denizr has joined #smartos[15:41:59] *** Tekni has quit IRC[15:48:02] *** Tekni has joined #smartos[15:55:59] *** denizr has quit IRC[15:57:20] *** neophenix has joined #smartos[15:58:35] *** denizr has joined #smartos[15:59:35] *** Z` has joined #smartos[16:01:39] *** Z`` has quit IRC[16:06:17] *** denizr has quit IRC[16:11:32] *** denizr has joined #smartos[16:12:00] *** Licenser has quit IRC[16:14:00] *** Z`` has joined #smartos[16:15:03] *** Tekni has quit IRC[16:15:45] *** rfvizarra has joined #smartos[16:16:46] *** Z` has quit IRC[16:17:34] *** KFR has joined #smartos[16:18:23] <KFR> rmustacc: e1000g driver from smartos worked great under kvm'd oi. My nfs works now.[16:20:25] *** xinkeT has joined #smartos[16:20:57] <killfill> Hi[16:23:49] <killfill> https://github.com/joyent/pkgsrc/tree/trunk/www/ap2-fastcgi i see there is that pkgsrc port (since 2 years ago), but ap2-fastcgi is not avaible as a package on http://pkgsrc.joyent.com/sdc6/2012Q1/x86_64/www/[16:24:01] <killfill> Why not?[16:24:31] <jperkin> not all packages are built for those repositories, we add extra features such as SMF but haven't got full coverage yet[16:25:01] <jperkin> an option is to use my full bulk build (http://www.perkin.org.uk/posts/9000-packages-for-smartos-and-illumos.html)[16:25:59] *** ira has joined #smartos[16:26:16] <killfill> jperkin: it seems that im locked in with apache-worker.. there is no official package for php-cgi nor apache-fastcgi, so cannot use php via fastcgi or fcgid or php-fpm.. :S[16:26:33] <killfill> (i know php it pretty boring.. bu need to use it.. :P)[16:27:03] <MerlinDMC> killfill, but isn't fpm part of php 5.3?[16:27:30] <jperkin> other options are to build it yourself (grab that pkgsrc tree, cd to that dir, 'bmake install'), or prod mamash to add it to our repo :)[16:27:43] <killfill> MerlinDMC, yup, but you need fastcgi to use it.. at least all docs on the web i could find configure it that way (with the FastCGIExternalServer thing)[16:28:31] <MerlinDMC> and the fpm binary is not inthe package ... so nevermind[16:28:55] <killfill> oh yes, there is php-fpm[16:29:26] <MerlinDMC> ah ... extra package ... k[16:29:40] <killfill> i guess php is not very poppular between joyent costumers.. :P[16:29:57] <MerlinDMC> i have to use that 8h/d ... i don't like it[16:30:00] <konobi> it is, it's just packaged differently than you might be used to[16:30:46] <killfill> konobi: with apache+prefork thing workd greate.. its just apache+worker works faster[16:31:18] <konobi> people still use apache?[16:31:26] <killfill> heh...[16:31:41] <MerlinDMC> konobi, some people have to -.- (me also)[16:31:58] <MerlinDMC> the 2. thing i hate about my regular workday[16:32:07] <killfill> jaj..[16:32:10] <killfill> me too.. :P[16:32:12] <konobi> mmmm... bloaty[16:34:12] * MerlinDMC driving home now[16:36:00] <Alasdairrr> There's always OpenWebstack[16:36:03] <Alasdairrr> http://wiki.smartos.org/display/DOC/3rd+Party+Software+Repos[16:38:13] <killfill> Alasdairrr: they dont have apache-fastcgi also.. :S[16:38:34] * killfill trying from http://pkgsrc.smartos.org/packages .. :)[16:38:59] <Alasdairrr> i think we do[16:39:01] * Alasdairrr checks[16:40:18] <Alasdairrr> https://hg.openindiana.org/users/aszeszo/s10-userland/rev/c6132ccfd9a6[16:40:19] <Alasdairrr> hmm[16:40:35] <Alasdairrr> looks like it may not be published yet[16:41:22] <killfill> :)[16:41:50] <Alasdairrr> let me see if i can get that published for you[16:42:24] <killfill> im trying with the url above.. the packages installed fine so far.[16:44:18] *** Sachiru has joined #smartos[17:13:13] *** mamash has joined #smartos[17:13:46] *** dap has joined #smartos[17:14:46] *** jdavis1 has joined #smartos[17:26:26] *** denizr has quit IRC[17:26:46] *** denizr has joined #smartos[17:34:48] *** Gurpartap has quit IRC[17:41:53] *** denizr has quit IRC[17:44:16] *** xinkeT has quit IRC[17:45:51] *** xinkeT has joined #smartos[17:50:33] *** denizr has joined #smartos[18:00:53] *** bens1 has quit IRC[18:02:45] *** xinkeT has quit IRC[18:03:11] *** xinkeT has joined #smartos[18:13:36] <siezer> hrmmm... does the root password for 20120713T005832Z not work for anyone else?[18:18:59] <nahamu> siezer: how did you boot?[18:27:25] *** denizr has quit IRC[18:27:46] *** denizr has joined #smartos[18:31:13] <MerlinDMC> killfill, tried ap22-fcgid from pkgin?[18:32:00] <killfill> MerlinDMC: yup, but it seemd to require a php-cgi[18:32:24] <killfill> MerlinDMC: im playing with ap22-fastcgi from 2012Q2[18:32:30] <killfill> (to used it with fpm[18:32:35] <MerlinDMC> killfill, depends on configuration ... php-fpm is usable as well[18:34:43] <siezer> nahamu usb key live (no install)[18:34:49] *** enmand has joined #smartos[18:44:58] *** jdavis1 has quit IRC[18:45:26] <nahamu> siezer: if you're not booting single user, I think the root password is "root"[18:46:00] <nahamu> (controlled by the hash on the kernel command line)[19:05:12] <rmustacc> So office hours are on as always.[19:07:12] <nahamu> I'm mostly back to working on spice stuff these days. We have QEMU 1.1.0 patched to run on SmartOS and the Spice devs fixed a bug that was causing crashes, so hopefully I'll have some useful stuff for other people very soon.[19:07:13] *** pijewski has joined #smartos[19:08:26] <nahamu> I'm hoping to be able to turn over the QEMU 1.1.0 stuff to Joyent at some point.[19:13:09] *** Sachiru has quit IRC[19:15:47] <nahamu> rmustacc: anything interesting coming down the pipe?[19:16:31] <rmustacc> In this week's release?[19:16:58] <nahamu> yeah[19:18:15] <rmustacc> You have the umem stuff, some work to improve scheduling when you are at your cap.[19:18:45] <Alasdairrr> umem stuff?[19:19:34] <rmustacc> Alasdairrr: I recently finished up some work on libumem to help improve allocation time in certain circumstances.[19:19:46] <rmustacc> Specifically for smaller, short-lived, repetative allocations.[19:20:22] <Alasdairrr> rmustacc: Ah hah, cool :-) is this with a view towards making it the default memory allocator?[19:20:37] <rmustacc> Certainly helps going down that path, eyah.[19:20:39] <rmustacc> *yeah[19:20:55] <KFR> Do people use smartos for desktop virtualization? I'm thinking Sunray/ORacle vdi[19:21:11] <nahamu> KFR: I'm working on Spice integration for that[19:21:18] <nahamu> http://wiki.smartos.org/display/DOC/Spice+on+SmartOS[19:21:30] <nahamu> I need to update that page with some new stuff[19:21:42] <nahamu> it's approaching what I would consider beta quality.[19:22:08] <nahamu> but I think other people may have gotten the Sunray software running on SmartOS[19:22:32] <KFR> Probably within KVM[19:22:38] <rmustacc> Alasdairrr: All the gory details are here: http://dtrace.org/blogs/rm/2012/07/16/per-thread-caching-in-libumem/[19:23:00] <Alasdairrr> cheers rmustacc *reads*[19:24:10] <KFR> Anyone know if virtualbox will run inside of kvm?[19:24:42] <MerlinDMC> rmustacc, "To try it out, grab the upcoming release of SmartOS and either add -lumem to your Makefile" ... that means the makefile of the software i build on smartos or the smartos release itself?[19:27:32] <nahamu> MerlinDMC: I think that refers to the Makefile of the software you want to use libumem.[19:28:54] <rmustacc> KFR: Nope.[19:29:08] <rmustacc> KFR: Well, actually, it can run. But the performance will be terrible.[19:29:37] <rmustacc> MerlinDMC: The software you want to run.[19:29:53] <rmustacc> You need to either explicitly link against it or LD_PRELOAD it.[19:34:24] *** Licenser has joined #smartos[19:34:56] *** rfvizarra has quit IRC[19:37:05] *** wdent has joined #smartos[19:37:28] <MerlinDMC> k thx[19:48:25] *** xinkeT has quit IRC[19:52:25] *** jdavis1 has joined #smartos[19:57:36] *** marsell_ has joined #smartos[20:01:00] *** marsell has quit IRC[20:01:01] *** marsell_ is now known as marsell[20:02:33] <bdha> marsell: Nice. That will be good.[20:05:37] *** xinkeT has joined #smartos[20:07:05] *** KFR has quit IRC[20:13:07] *** xinkeT has quit IRC[20:13:24] *** xinkeT has joined #smartos[20:55:03] <killfill> how can i find out what process is using my swap?..[20:55:52] <Meths> prstat[21:01:25] <richlowe> ps -eo pid,comm,vsz | sort -nk3 perhaps, also.[21:22:01] *** knweiss has joined #smartos[22:31:19] *** knweiss has quit IRC[22:51:09] *** denizr has quit IRC[22:51:27] *** denizr has joined #smartos[22:54:26] *** Meths has quit IRC[22:55:51] *** MacSpengo has joined #smartos[22:56:58] <MacSpengo> How does one go about using NFS storage to store my VMs created in SmartOS?[22:57:38] <rmustacc> Generally, you don't.[22:58:06] <MacSpengo> so local storage is the only way?[22:58:26] <rmustacc> For the default backing store for zones and KVM instances, yes.[22:58:33] <rmustacc> You can do NFS mounts inside them, but the core is all on local storage.[23:00:32] *** Meths has joined #smartos[23:00:50] <MacSpengo> Keep the VM disk small to install the OS and then use NFS/NAS/SAN storage to store the data within the VM?[23:01:36] <rmustacc> If that's what you want.[23:02:39] <MacSpengo> Makes sense, this way it is easier to move the VM to a different node if you need to.[23:03:26] <MacSpengo> I need to store right around 20TB. I was going to create the VM disk size at 20TB. But thought I would ask if there was a better way to do it.[23:04:18] *** enmand has quit IRC[23:09:42] <MacSpengo> rmustacc: thanks for your help[23:10:08] <rmustacc> Creating a single 20 TB volume should be fine.[23:10:37] <rmustacc> Assuming your local storage is appropriately using RAID and the like.[23:12:32] <MacSpengo> Yes, but if I ever want to move that VM to a new piece of hardware. I have to move that whole 20TB.[23:13:03] <rmustacc> That's true.[23:13:20] <rmustacc> If you're NAS goes down you lose everything.[23:13:21] <MacSpengo> That is why I was wondering if there was a way to use shared storage to host the VMs in smartos[23:14:21] <MacSpengo> Yes, but I can setup a couple of NASs using OI and setup ZFS replication to try and keep my data safe[23:15:40] <MacSpengo> Can the global zone of SmartOS be configured to be an NFS server?[23:16:29] <rmustacc> Yes, it can be.[23:17:40] <MacSpengo> rmustacc: Then that is the direction I think I will go with. Setup a machine to be a NFS server with my 20TB and then setup up a machine to run my VMs.[23:18:13] <MacSpengo> This way I can also use the storage(NFS)system to run the VM if I need to do anything with the other machine.[23:20:42] <benbangert> MacSpengo: I tried OI first, it had hardware issues, and the QEMU/KVM stuff was messed up[23:21:49] <benbangert> I'm using the ZFS NFS thing from the GZ[23:22:00] <benbangert> there's a recipe on the SmartOS on maintaining users/groups in the GZ across reboots[23:23:49] <MacSpengo> benbangert: I have heard that. I think I will setup Smartos on the storage system as an NFS server and then my second machine with SSDs to run the VMs.[23:25:16] <MacSpengo> benbangert: then I can run the VMs temporarily on the storage system if I need too.[23:25:28] <benbangert> I run the VM's on the same machine[23:25:50] <benbangert> they're decently fast, though I'll prolly add a separate mirror SSD ZIL and/or a L2ARC SSD for better performance[23:26:55] *** xinkeT has quit IRC[23:27:37] <benbangert> right now everything stays in RAM enough[23:27:38] *** xinkeT has joined #smartos[23:34:51] *** enmand has joined #smartos[23:38:41] <MacSpengo> benbangert: I am moving and changing mass amount of data per day. So I will setup the storage system with 10Gbe to NFS to the VM running on a node using 10Gbe.[23:39:28] <rmustacc> You want to drive 10 GBe to a KVM instance?[23:39:33] <MacSpengo> benbangert: This should allow the storage to run great and then VMs to run great on another machine with SSDs hosting the VM disk for the OS[23:40:57] <MacSpengo> rmustacc: Sure, this what I am already doing with VMWare and it works great. I just looking for ways to get away from the VMWare stuff and really start building a cloud using smartos[23:41:58] <rmustacc> Well, I can tell you that you won't be getting 10 GB/s to a KVM VM.[23:43:00] <MacSpengo> rmustacc: Your right, just looking for speeds better than gigabit. Right now the KVM VM I am testing with is getting about 355MB/s[23:43:31] <rmustacc> Great.[23:43:40] <rmustacc> Better than I usually see.[23:43:40] <MacSpengo> After I get all of this figured out, I will move it into production[23:44:26] <MacSpengo> rmustacc: have you tried 10Gbe to a KVM VM? What speed are you getting?[23:45:04] <MacSpengo> rmustacc: I was getting about 80MB/s or so with gigabit and jumbo frames[23:46:47] <MacSpengo> rmustacc: this is using OI for the storage system and a smartos machine with 4 128GB SSDs for the OS drive of the VMs[23:51:18] *** echelog-1 has joined #smartos