July 6, 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 6, 2012 [00:00:27] <ryancnelson> right. so, if we've got a 96GB bare metal machine, and you buy an 8gb smartmachine, you've bought 1/12th of that machine's fixed resources. … we'll give you shares in that range, so you're guaranteed to get what you've paid for. (the formula is your shares / total shares provisioned… so the numbers don't have "units" ) … but we'll cap you somewhere higher than that[00:00:51] <ryancnelson> just so there's motivation to buy what you need, and not just hope to be cpu-bursting all the time[00:01:06] <killfill> greate[00:01:39] <ryancnelson> in your cloud, if you have 10 zones, each with 666 shares, they're all the same.[00:01:51] <ryancnelson> one with 333 will get half the juice[00:01:58] <killfill> there is just 1 think im missing tho... on the packages setup on SDC, i can only setup the CPU Cap, whats the name of the cpu share?[00:02:52] <ryancnelson> shares in sdc are auto-calculated, so you don't oversell. it's based on ( your-ram / total-ram )[00:03:01] <ryancnelson> since ram is the finite resource[00:03:15] <killfill> aah[00:04:18] <ryancnelson> (you can't just dial all the shares up to 11, or they don't mean anything anymore, since they're relative to each other. so SDC handles it)[00:04:49] <ryancnelson> but in solaris-land, shares are getting set. in smartos.org stuff, if you make zones by hand, you set that.[00:07:12] <killfill> i think this is the reason why moving smartmashine's between compute nodes is not something you guys normally do. You buy HW, design how many VM will be on it.[00:07:41] <killfill> and not try to push every new VM's on a box that has 1% resource free[00:07:52] <killfill> like most datacenters do (at least the ones i know...)[00:07:59] <killfill> (i guess including amazon?..)[00:09:37] <killfill> or maybe im too far away from the truth.. :P[00:10:58] *** neophenix has quit IRC[00:12:35] *** trentm has quit IRC[00:12:39] <ryancnelson> well, we don't let people pick where they get provisioned (nobody does) … but we strive to put 'em where they'll be happiest, at the time… while not provisioning someone where they'll spoil a nicely sized "hole" that would accommodate a new 16GB customer.[00:12:42] <ryancnelson> it's like tetris[00:13:20] <ryancnelson> people can't say "i'd like a new 1GB vm on an empty box please"[00:15:12] *** denizr1 has joined #smartos[00:17:14] *** denizr has quit IRC[00:18:12] * killfill trying to get it[00:19:28] *** ohlol has joined #smartos[00:22:38] <killfill> ryancnelson: what there is a box that has 90% used resources, and comes a client that need 20% of that, if you set it in the same mashine, the shared of everyone else will lower, right?..[00:22:56] <killfill> what there/ if there[00:23:19] <ryancnelson> 90% of cpu currently used?[00:23:44] <ryancnelson> not 90% of the machine provisioned….[00:23:51] <killfill> 90% of "cpu shares" like you have a box with 10G, and 9 clients with 1G 'share'[00:24:11] <rmustacc> Your shares are based off the total number of shares on the system.[00:24:19] <ryancnelson> oh… then a new customer who wants 2 "shares" doesn't fit on that box[00:24:23] <ryancnelson> we won't oversell[00:24:26] <rmustacc> e.g. you've decreased the value of everyone elses by issuing more that way.[00:24:45] <ryancnelson> and SDC won't let you… which is why you can't set the shares explicitly in the package definition.[00:24:49] <ryancnelson> it's calculated.[00:27:07] <killfill> hm.. how does SDC not let you? i mean.. there is no way to see (i.e. in the operator panel) how many shares are user/avaible? it calculates it by ram?[00:27:39] <richlowe> letting people mess around with the shares calculations by hand is a recipe for disaster in that environment, I would imagine[00:27:41] <e^ipi> it calculates based on the total number of shares[00:28:16] <richlowe> given, as above, it's a purely relative calculation, you can't pin the value of just _one_[00:29:12] <rmustacc> Shares are based on DRAM.[00:29:16] <rmustacc> You can't set them.[00:29:17] <rmustacc> Only caps.[00:29:40] <killfill> ok[00:38:46] *** denizr has joined #smartos[00:40:21] *** denizr1 has quit IRC[00:41:20] <ryancnelson> if you rent half the box's ram, you get half the box's shares.[00:41:38] <ryancnelson> … and so on.[00:42:58] <ryancnelson> … if all your neighbors are asleep, you can borrow a cup of CPU from them, but when they wake up, they get it back.[00:43:24] <killfill> so what clients of the same box share is cpu, not ram. Sum of all ram of the clients should be <= total ram of the box.. not so the cpu[00:43:42] <ryancnelson> right. you don't share ram[00:44:09] <ryancnelson> … and actually, we reserve some ram for us, too.[00:44:23] <ryancnelson> so we don't sell 96 1GB vms on a 96gb box[00:44:33] <killfill> hehe... sure :)[00:44:44] <ryancnelson> that would be overselling, too, and would kinda suck[00:46:02] <ryancnelson> but yeah… so you understand why letting shares be tweakable is a recipe for another customer not getting what they paid for. and being able to tell. so we don't do it.[00:46:28] <killfill> yup[00:47:04] <killfill> "if you rent 32 of a 64G box, you will get a cpu share of the half of the cpus"..[00:49:12] <ryancnelson> yep… and back to your original question about migrating customers and stuf… if you ask for us to resize that up to a 64gb machine, and there's other people already there, we can't/won't do it.[00:49:22] <ryancnelson> 'cause there's not room.[00:49:37] <ryancnelson> (as opposed to taking your money and smiling :)[00:49:45] <killfill> heh..[00:49:48] <killfill> yup i get it.[00:50:45] <killfill> so if i create a 2G mashine on my 16G box, and have it 'cpu capped' to 100, if the mashine has 8 cpus, i should get the performance of 1 cpu?[00:51:33] <ryancnelson> yeah… pretty close, anyway.[00:51:40] <ryancnelson> … but you'll see all 8 cores[00:52:00] <ryancnelson> … so you can be more parallel with your workload, in theory[00:52:17] <killfill> thats the wired think.. as far as i see, i get much more that 1 cpu performance[00:52:36] <ryancnelson> are you actually capped?[00:52:43] <killfill> but i have free resources, i guess its becouse of that[00:52:49] <killfill> yup[00:53:02] <ryancnelson> that's a good way to tell...[00:53:34] *** Orpheus has quit IRC[00:53:42] <ryancnelson> redline it with some benchmark that uses multiple threads… and you should still see the right amount of idle on the global zone's cpu[00:55:23] <ryancnelson> killfill, where are you, geographically?[00:55:33] <ryancnelson> brazil?[00:55:47] <killfill> nope, from Chile[00:56:02] <ryancnelson> ah. 200.X.X.X looked south american :)[00:56:21] <ryancnelson> i used to spend a lot of time there[00:56:34] <killfill> oh really? cool[00:57:28] <killfill> here in chile?[00:58:26] <ryancnelson> i spent 12 hours in santiago once, for a meeting… the rest was usually in SP, rio, or BA[00:58:52] <ryancnelson> do you remember "starmedia"?[00:59:00] <killfill> meeting related to joyent?[00:59:35] <ryancnelson> not joyent… this was in the 90's, i worked at a latin american portal site[01:01:29] <killfill> ryancnelson: ooh cool.[01:02:02] <killfill> ryancnelson: i actually work on betazeta.com, wish kind of look siilar of what startmedia is/was :P[01:02:57] <ryancnelson> yeah, they were trying to be yahoo… and ended up being what yahoo is now… a giant mess :)[01:03:02] <ryancnelson> ahead of their time![01:03:35] <killfill> jaja[01:04:14] <ryancnelson> i was a millionaire for 6 days … now i'm a hundred-aire… :)[01:04:44] *** Orpheus has joined #smartos[01:27:39] *** andoriyu has joined #smartos[01:35:57] *** dap has quit IRC[01:52:02] *** mamash has left #smartos[02:26:57] *** scanf has joined #smartos[02:47:30] *** pijewski has quit IRC[02:55:06] *** ryancnelson has quit IRC[02:58:00] *** jdavis has quit IRC[03:04:30] *** darjeeling has quit IRC[03:27:58] <jeffpc> does smartos somehow remember to not use kvm if it didn't attach at boot?[03:29:42] <jeffpc> ah, there was no /dev/kvm[03:59:53] *** darjeeling has joined #smartos[04:06:38] *** andoriyu_ has joined #smartos[04:08:30] *** andoriyu has quit IRC[04:29:41] *** dis has quit IRC[04:31:37] *** dis has joined #smartos[04:52:01] *** denizr has quit IRC[05:13:27] *** darjeeling has quit IRC[05:32:29] *** marsell has quit IRC[05:40:23] *** darjeeling has joined #smartos[05:40:57] *** ira has quit IRC[06:08:30] <ismell> ohh botting the vmware version[06:54:36] *** sam_wan has joined #smartos[07:11:39] *** andoriyu_ has quit IRC[07:33:10] *** dis has quit IRC[07:45:54] *** mrb has joined #smartos[07:52:08] *** Licenser has quit IRC[08:00:46] *** dis has joined #smartos[08:02:29] *** mrb has quit IRC[08:11:47] *** sam_wan has quit IRC[08:13:25] *** dis has quit IRC[08:15:05] *** sam_wan has joined #smartos[08:25:42] *** dis has joined #smartos[09:00:42] *** Licenser has joined #smartos[09:00:55] *** marsell has joined #smartos[09:16:43] *** AikoM has joined #smartos[09:22:57] *** mamash has joined #smartos[09:36:13] *** bens1 has joined #smartos[10:25:36] *** jamesd has quit IRC[10:51:17] *** darjeeling has quit IRC[11:14:26] *** sam_wan has left #smartos[11:17:16] *** Licenser has quit IRC[11:20:50] *** marsell_ has joined #smartos[11:23:13] *** marsell has quit IRC[11:23:13] *** marsell_ is now known as marsell[11:47:08] *** Licenser has joined #smartos[12:32:23] *** jeekl has quit IRC[12:32:30] *** jbergstroem has quit IRC[12:45:16] *** jeekl has joined #smartos[12:49:26] *** jbergstroem has joined #smartos[12:57:43] *** bens1 has quit IRC[13:29:01] *** marsell has quit IRC[13:54:56] *** JoshWilsdon has quit IRC[14:00:52] *** Licenser has quit IRC[14:01:48] *** sebasp has quit IRC[14:04:01] *** ira has joined #smartos[14:05:11] *** sebasp has joined #smartos[14:11:01] *** v12 has quit IRC[14:11:18] *** v12 has joined #smartos[14:13:24] *** marsell has joined #smartos[14:43:11] *** Licenser has joined #smartos[15:14:05] *** bens1 has joined #smartos[15:23:07] *** max` has joined #smartos[15:31:57] *** CarlosC has joined #smartos[15:38:55] *** enmand_ has joined #smartos[15:41:16] *** elijah-mbp has joined #smartos[15:45:52] *** elijah-mbp has quit IRC[16:06:09] <killfill> i setup sdc to use a ntp server with its dns name.. as it doesnt has dns enabled.. where do i change it to use the ip?.. cannot find the config file. :S[16:08:05] <CarlosC> are you on the headnode?[16:08:28] <killfill> yup[16:08:48] <CarlosC> run the script: /usbkey/scripts/mount-usb.sh[16:09:00] <CarlosC> this will mount the usb key in rw to /mnt/usbkey[16:09:10] <CarlosC> modify the file: /mnt/usbkey/config[16:09:18] <CarlosC> umount /mnt/usbkey[16:09:22] <CarlosC> reboot headnode[16:09:29] <killfill> aaah greate thanks![16:09:35] <CarlosC> no worries[16:10:05] <CarlosC> btw…the usbkey is present in /usbkey but it's in read only[16:10:39] <CarlosC> makes it easy to see what the config is without being worried about messing with it[16:11:13] <killfill> yah i get it.. well it seem to by rw, but i guess if i only change that in the instance, not in the usb drive[16:11:27] <CarlosC> yep…in memory[16:11:35] <CarlosC> not persistant to the usbkey[16:14:13] <killfill> greate[16:16:14] *** jamesd has joined #smartos[16:21:19] *** neophenix has joined #smartos[16:48:19] *** denizr has joined #smartos[17:22:59] <nahamu> wesolows: heads up, you might be the person with the answer to my question on the mailing list.[17:27:17] *** AikoM has quit IRC[17:29:38] *** ryancnelson has joined #smartos[17:40:41] *** jdavis has joined #smartos[17:54:01] *** darjeeling has joined #smartos[17:56:20] *** jamesd has quit IRC[17:57:05] *** jamesd has joined #smartos[18:04:10] <nahamu> hmmm... even after installing nodejs, my build still failed... :-/[18:05:48] *** ryancnelson has quit IRC[18:26:38] *** dap has joined #smartos[18:31:44] <wesolows> how did it fail? ronn.js is one of the last phases[18:35:40] <nahamu> I don't remember. I kicked off another one inside of typescript to generate a log file. it was somewhere in illumos-extra. may have indeed been ronn.js[18:46:41] <nahamu> uh, seems to have worked this time... false alarm?? I'm confused, but the build looks to be okay.[18:47:17] *** max` has quit IRC[18:49:33] <wesolows> ok. if we're hitting more unreproducible failures, it's probably because something somewhere doesn't like -j[18:51:04] <nahamu> silly make.[18:52:54] *** Licenser has quit IRC[18:55:23] *** Licenser has joined #smartos[18:56:12] *** denizr has quit IRC[18:56:34] *** denizr has joined #smartos[19:11:02] *** bens1 has quit IRC[19:11:28] *** jamesd_ has joined #smartos[19:14:12] *** jamesd has quit IRC[19:15:04] *** deirdres has quit IRC[19:32:24] <jeffpc> what's the reasoning for not having zones list any zfs datasets?[19:33:44] <nahamu> jeffpc: as in, if you don't explicitly delegate a dataset to a zone, why doesn't "zfs list" show the zones "root" filesystem?[19:33:53] <jeffpc> yeah[19:34:07] <nahamu> (My guess is because the zone filesystem contains more than just the zone's root filesystem...)[19:34:10] <jeffpc> I guess that's lofs's doing[19:34:27] <jeffpc> no wait, that's /usr whatever[19:34:54] *** dis has quit IRC[19:35:02] *** dis has joined #smartos[19:35:21] <nahamu> it might be interesting to consider a world where zones/<uuid> always contained a zones/<uuid>/root child filesystem that was always delegated.[19:36:33] <jeffpc> right[19:37:02] <jeffpc> IIRC, sol11 does some zfs name munging so the zone's root fs shows up as "rpool"[19:37:39] <jeffpc> (FWIW)[19:41:53] <nahamu> as to the reasoning, I have no idea.[19:42:50] <nahamu> there is certainly support in vmadm for adding an additional filesystem that *is* delegated.[19:45:35] <jeffpc> I'll have to play with that[19:46:59] <nahamu> look for delegate_dataset in the vmadm man page.[19:47:59] <jeffpc> will do[19:55:32] *** jelmd has joined #smartos[19:55:42] <Licenser> jeffpc had any luck?[19:57:09] <jelmd> has anybody running smartos on a E2600 based SM MB?[19:59:30] *** enmand is now known as enmand[i][20:00:36] *** enmand_ is now known as enmand[20:02:35] *** ziomatto has joined #smartos[20:06:44] <ziomatto> hi, can I install smartos on linux kvm ?[20:06:47] <jeffpc> Licenser: sorry, I got distracted with trying out smartos on real hardware :)[20:07:03] <jeffpc> ziomatto: sure[20:07:30] <Licenser> oh cool that is more important for sure :D[20:07:40] <ziomatto> jeffpc: thank you.[20:08:34] <jeffpc> ziomatto: you won't probably have any luck with kvm zones on smartos[20:09:20] <ziomatto> jeffpc: is vitualbox a better solution ?[20:09:32] <jeffpc> hell no[20:09:59] <jeffpc> AFAIK, vmware player for windows does it[20:10:23] <ziomatto> jeffpc: have to try smartos on bare metal to take advantage of all his feature ?[20:10:24] <jeffpc> but if you just want to experiment with smartos zones, you don't need nested virtualization[20:10:49] <jeffpc> ziomatto: yes, and no. I'm only scratching the surface of it[20:12:11] *** deirdres has joined #smartos[20:12:22] <rmustacc> ziomatto: If you need to run operatings systems that aren't SmartOS, then it won't work.[20:12:33] <rmustacc> If you don't need our KVM implementation, then you can run it under linux kvm all you want.[20:13:39] <ziomatto> rmustacc: ok. thank you.[20:14:13] <jeffpc> is SmartOS the guest or the host or both?[20:16:06] <ziomatto> smartos would be the guest[20:16:38] <ziomatto> it's only for testing.[20:17:29] <rmustacc> There's no virtio support though.[20:18:32] <ziomatto> a little slow[20:19:08] <jeffpc> rmustacc: did anything happen with the apix crashes?[20:19:23] <rmustacc> jeffpc: We disable the apix and are slowly trying to unravel that.[20:19:24] <jeffpc> rmustacc: or does it work for joyent -> not high enough priority?[20:19:28] <jeffpc> ah[20:20:05] <rmustacc> The problem is that figuring out any state after you lock up is hard.[20:20:31] <jeffpc> yeah[20:20:33] <rmustacc> Since you know, the apic is stuff at IPL 15.[20:21:31] * jeffpc is still rather unfamiliar with IPL, but knows that it's all tricky code[20:31:37] *** ziomatto has quit IRC[20:39:17] *** denizr has quit IRC[20:40:35] <sebasp> Is http://wiki.smartos.org/display/DOC/Building+Packages page up to date enough to use it?[20:41:00] <sebasp> I need to update some critical package for me and I want to make sure I do it the right way ![20:41:22] <rmustacc> sebasp: I think so.[20:41:55] <sebasp> rmustacc: Thanks… I'm not 100% used to pkgsrc, so I want to make sure I do it the right way :)[20:42:55] <rmustacc> No worries, neither am I. jperkin and mamash are your guides around there.[20:44:46] <sebasp> Also, are there any plans to have a "smartos" repo of packages? Not that joyent's stuff is not good, but something where we "the community" could contribute back…Kind of the /release /pending /contrib model of IPS but for smartos?[20:44:58] <sebasp> I will take note of the name...[20:45:10] *** denizr has joined #smartos[20:45:20] <rmustacc> sebasp: iirc jperkin was working to make the full pkgsrc dataset available in some capacity.[20:46:55] <rmustacc> sebasp: At which point, I guess the easiest way of adding packages is to add them there, or they may already be there.[20:47:16] <rmustacc> sebasp: Or hopefully it'll be easy to just go and add multiple repos for packages so that'd be easier to create.[20:49:33] <sebasp> rmustacc: Will start by working on https://github.com/joyent/pkgsrc.git , I can always open tickets on this and do pull request. And then run my own repo until it's built for the joyent repo.[20:52:26] <rmustacc> sebasp: Sure, but if the package is in upstream pkgsrc then you shouldn't need to do all the work, it's just a matter of getting it moved in there.[20:52:37] <rmustacc> We don't put everything in there, becuase not everything has things like SMF manifests, etc.[20:53:21] *** rahim_ has quit IRC[20:54:26] <rmustacc> sebasp: but I'm not the best resource as to plans with packaging in general, sorry.[20:56:22] <sebasp> rmustacc: No problem...[20:56:54] <sebasp> rmustacc: Anything is better then nothing…[21:05:06] *** CarlosC has quit IRC[21:23:12] *** sebasp has quit IRC[21:38:26] *** sebasp has joined #smartos[21:55:49] <sebasp> deirdres: smartos.org css needs some rework :) http://smartos.org/tag/jeff-bonwick/[22:02:48] *** sebasp has quit IRC[22:10:48] *** neophenix has quit IRC[22:15:48] <richlowe> doh.[22:21:48] *** Licenser has quit IRC[22:25:07] *** enmand has quit IRC[22:25:55] <deirdres> sbasp: sigh, thanks. Thought I asked our web guy to fix that two months ago[22:36:41] *** deedubs has joined #smartos[22:45:53] *** bens1 has joined #smartos[22:46:37] *** dis has quit IRC[22:47:01] *** bens1 has quit IRC[22:57:40] *** enmand has joined #smartos[23:01:46] *** ira has quit IRC[23:06:19] *** deedubs has quit IRC[23:24:36] *** luuse has joined #smartos[23:28:30] <Saskaloon> Is there pci-passthrough, with KVM virtualization?[23:29:05] <konobi> nope[23:31:14] <Saskaloon> I thought the Linux implementation of KVM supported it; but, I may be confusing things with Xen.[23:32:02] <rmustacc> Yes, Linux KVM supports it. We did not implement it.[23:32:19] <Saskaloon> So, it could be possible with further development.[23:32:26] <rmustacc> Yes.[23:32:35] <rmustacc> Nothing we did preculdes adding that functionality.[23:32:45] <rmustacc> We are currently unlikely to be the ones to do it however.[23:33:05] <rmustacc> Really the only case where it makes sense is with GPUs right now and we don't sell GPU access at Joyent.[23:33:15] <Saskaloon> Interesting... I'll have to see what is required for it.[23:33:16] <rmustacc> At least, in our environment.[23:33:26] <rmustacc> For PCI passthrough?[23:33:51] <rmustacc> A bunch of changes to the driver and probably some way of stopping or detaching illumos from using it.[23:34:44] <Saskaloon> Yeah. I was debating virtualizing a MythTV linux box; but, I have doubts about whether PCI-passthrough would work, on a TV tuner, if the illumos kernel doesn't know about the hardware.[23:35:13] <rmustacc> PCI-passthrough would work even better if the kernel doesn't have knowledge about the deviec.[23:35:17] <rmustacc> *device[23:35:31] <rmustacc> The general challenge with PCI-passthrough is getting the host kernel to not pay attention to it.[23:38:28] <Saskaloon> Which if illumos doesn't speak to it, and if the kernel doesn't disable the pci slot - thinking it's problematic hardware, then pci-passthrough and letting the VM manage the slot directly may actually work.[23:39:05] <rmustacc> Well, if our kvm implementation supported VT-d, yes.[23:39:24] <Saskaloon> The alternative would be to investigate the current state of migration attempts of the MythTV components.[23:39:52] <Saskaloon> Ah, so that would be one hurdle, for sure.[23:41:57] <rmustacc> Yes, and a rather large one, in all honesty.[23:42:20] *** Licenser has joined #smartos[23:43:08] <Licenser> dun dun dun[23:45:41] <Saskaloon> Well, I had put together a Xeon workstation, with VT-d support in the cpu and on the mobo. This large software component could theoretically open-up the door to some new functionality in the guests.[23:46:19] <Saskaloon> I'm guessing I'd need to find the Linux sources, some documentation from Intel, and a lot of time. :)[23:47:27] <rmustacc> Yup, there's a lot of time.[23:47:39] <rmustacc> But, there are a few things you can do to help get you started[23:47:50] <Licenser> nested virtualisation?[23:47:52] <rmustacc> First off, the Linux branch that we started on is sitting in the kvm repo itself.[23:48:04] <rmustacc> Licenser: Nah, something at least marginally useful VT-d.[23:48:25] <Licenser> heh nestedvirtualisation is usefull[23:48:40] <Licenser> You could make a cloud in a cloud![23:48:51] <rmustacc> Sure, but performance.[23:48:55] <Licenser> like amazon could run AWS inside of Joyent[23:48:57] <Saskaloon> How would you nest virtualization? A nested hypervisor, sharing with the hardware one? (Hypervisor scheduler???)[23:49:33] <Licenser> Saskaloon pretty much the same way you virtualize a normal OS[23:49:47] <Licenser> just that there need to be some trickery for hardware virtualisation[23:50:02] <konobi> aws is xen based[23:50:19] <e^ipi> level 3 tries to do some priveleged op, which is caught by level2 hypervisor, which does it's work to prepare it, then does a hardware call which is caught by level 1, which does some work, and then actually calls hardware[23:50:22] <Licenser> konobi so you can run AWS in XEN in JoyentCloud![23:50:23] <e^ipi> *shrug* s[23:50:27] <rmustacc> You need to emulate the hardware instructions that make up VMX.[23:50:35] <e^ipi> seems pretty straightforward (conceptually)[23:51:18] <Licenser> Well for me it would be nice since I could test fifo in a fully virtualized SmartOS without the reisk of braking my server :P[23:52:10] <rmustacc> You could do that already, modulo running kvm.[23:52:15] <Saskaloon> I'll be back-and-forth a bit, shortly here. So, what would be the ideal development platform for KVM development?[23:52:18] <Licenser> but it's not really nessessary in the end the only think not testable outside vmware is kvm[23:52:22] <Licenser> *inside[23:52:35] <Licenser> rmustacc *nods*[23:52:44] <e^ipi> Saskaloon: something with vmx. you can put together a box for < $200[23:54:08] <rmustacc> Saskaloon: You'll want something with a serial console and with ipmi.[23:54:25] <rmustacc> Being able to send an nmi when you inevetibly deadlock the kernel is pretty useful.[23:54:40] <rmustacc> Serial console probably slightly less useful.