[00:33] *** toorop has quit IRC (Ping timeout: 248 seconds) [00:40] *** toorop has joined #arpnetworks [00:42] *** Ehtyar has quit IRC (Quit: +++ OK ATH OK) [01:07] *** arenlore has joined #arpnetworks [01:09] *** arenlor has quit IRC (Ping timeout: 276 seconds) [05:49] *** Webhostbudd_ has joined #arpnetworks [08:29] mercutio: you hailed [08:46] In just the last couple days, we've received two orders from IPv6 addresses. This is unprecedented. IPv6 is starting to take! [08:48] lol [08:58] it's not a laughing matter ;) [08:58] up_the_irons: Nitrous Oxide is no laughing matter. [08:59] ... [08:59] up_the_irons: Too painful for you? [09:00] ___ [09:05] sounds like dark matter now [09:27] would be nice if I thought I'd ever get ipv6 from my ISP [09:27] and that if I did, it was more than a single address :P [09:42] does anyne have any recommendations for a windows vps host? [09:42] * ryk shudders [09:43] I can put a Windows ISO into a VPS' CD-ROM if u like... (and I will feel dirty doing so too) [09:43] aside from amazon [09:43] up_the_irons: have you tested it? [09:44] ryk: no [09:45] it would be fine for me, if i wanted windows. [09:45] but i need to recommend something good to a client [09:45] ah ok [10:09] up_the_irons: you'll probably think this is crazy talk [10:09] but it might be a good idea to get a windows datacenter edition and offer that to clients [10:09] i have a SMB customer who wants to move their whole internal infrastructure to Amazon [10:10] I am trying to find something to recommend to them other than AWS, but can't really find anything that offers private vlans [10:13] there are only a few windows vps providers out there, and none of them seem to allow you to create a private network. [10:14] ryk: HMMMMMMMMMMMMMMMMMMM.... [10:14] while also allowing the VPS's to be on different hosts. [10:14] ryk: do you have a link to the license I need? does your SMB customer have their own licenses? [10:15] ryk: yeah, my infrastructure totally lets you do that :) [10:15] I think they have their own licensing, yes. [10:15] it is actually a customer of a customer. :) [10:16] ryk: see, if they have their own licenses, I believe it is legal for me to pop a Windows ISO onto a blank VM and let them install with their own license [10:16] ryk: do you know how many hosts we're talkin' about? [10:16] there actually is a dedicated virtual license that microsoft has [10:16] just two hosts, so they can have some level of redundancy [10:19] ryk: their data center license seems to be tied to Hyper-V [10:19] i wish i knew more about it. [10:19] trying to read up myself now. [10:19] according to this, http://www.michaeldolan.com/1034, Win Server 2008 runs fine on KVM/QEMU, and this post is way old... :) [10:20] yeah i know of another kvm provider that lets customers do it [10:21] ryk: which one? [10:21] BlueVM [10:21] they are one of the cheapie reseller types [10:21] ah [10:22] I really wouldn't trust any important infrastructure on them. [10:23] ryk: i'd be happy to let them trial my setup. I've ran Win 2008 on KVM on a customer's host once, so i know it works [10:23] Ok. well that might be a good option then. [10:24] What raid level do you have again on your hosts? [10:24] you might want to figure out how to indemnify yourself against licensing violations :) [10:24] ryk: 8x RAID 10 [10:29] i am almost positive that the datacenter edition does not require you to use the hyper-v hypervisor in order for the 'unlimited instances' licensing to be in effect. [10:30] here is something old that suggests that: http://www.vmware.com/pdf/ms_licensing_faqs.pdf [10:32] this is what you would need, i think. http://www.provantage.com/microsoft-p71-06484~7WWFF6M9.htm [10:32] if you were to offer it on your licensing. [10:33] i don't know how the CALs would be handled, though. [10:34] btw up_the_irons i discovered that NexentaStor commercial licensing has software that can do an automatic failover between the SuperMicro Twin SBB two-node hardware. [10:38] ryk: oh nice [10:40] ryk: is that like a san? [10:41] exactly. what i was talking about before. [10:41] ryk: wow, $3600 for that data center license. i'd have to sell a lot of VMs to make that worth it [10:41] yeah :/ [10:41] but the smb type customers should already have their own licensing [10:41] fuck that.. BYOL (bring your own license :) [10:42] just put it in your terms [10:43] yeah [10:44] what i eneded up proposing for our san this: [10:45] two of these (one a spare) for the front-end: http://www.siliconmechanics.com/i28372/single-xeon-1U.php [10:45] the HBA: http://www.newegg.com/Product/Product.aspx?Item=N82E16816118147 [10:45] the JBOD canister: http://www.amazon.com/gp/product/B004IY1N32/ [10:45] two of each of those, too [10:45] plus disks, switches, etc [10:45] and just do the manual failover thing. [10:51] ryk: oh, very nice. silicon mechanics is expensive though. try abmx.com [10:52] i did, but they didn't have the ratios of hardware i wanted [10:52] ryk: ratios? [10:52] well i have measured my VM resource usage internally here [10:52] right now I am sucking up memory but wasting CPU [10:53] so I had some particular specs requirements in order to get the best value [10:53] plus I need lots of NICs. that one I liked to, it has two PCIe, so I could put in a 4-port NIC, plus the 4-port on the mobo, plus the HBA [10:54] ryk: if you just call abmx and tell them that's what you want, they'll build it [10:54] i only put the links in because you asked at one point [10:54] they are pretty easy to work with [10:54] didn't know, thanks! [10:54] :) [10:54] ryk: yeah, tnx for the links too [10:54] ryk: nice canister; cheap, simple [10:54] this is for ZFS. You would want actual RAID cards instead of HBA for Linux storage. [10:55] yeah, there is a model with dual power supplies [10:55] ryk: yeah was gonna say, no dual power?! [10:56] but a friend of mine had a jbod with dual power supplies, and the part that went out on Christmas day was the part between the disks and the power supplies, that was not redundant. [10:56] ryk: holy.shit. [10:56] it was an electronic component [10:56] yeah so fuck that, just get two of em :) [10:56] the power bus, i assume [10:56] ok, i didn't know what it was called [10:56] ryk: that's always what i fear about single chassis things [10:57] much better to just have two [10:57] the one with dual power supply cost 2x as much anyway [10:57] ryk: so the manual failover, is that part of nexentastor? [10:57] http://www.amazon.com/DS-24DR-External-Rackmount-Enclosure-Redundant/dp/B007Z5TE9G/ [10:58] no, nexentastor has a plugin that gives you HA/failover [10:58] it's commercial. not part of the free edition. [10:58] I'm not going with that. [10:58] it seems to have been specifically written for the supermicro SBB [10:58] ryk: ah ok [10:59] this thing. http://www.supermicro.com/products/nfo/sbb.cfm [10:59] ryk: how many disks can be put on one SFF-8088 cable? [10:59] where each node can can access the disks [10:59] tons. [10:59] i don't even know. it expands out in the jbod [10:59] it's in the specs there but i have to go [10:59] ryk: ok, tnx for the info, later [11:00] as far as sans go, nexenta's pricing is extraordinarly reasonable for the high availability. http://nexenta.com/corp/store?page=shop.browse&category_id=3 [11:03] ryk: as far as redundant power goes though, while it can't protect against the power bus failing, what you really want is to protect against the UPS, panel, power strip or otherwise circuit failure. it's more of an external failure thing [11:24] or the psu itself failing. [11:39] yeah [11:51] *** arenlore is now known as Arenlor [12:06] *** Arenlor has left [12:17] up_the_irons: is there a network issue by any chance? [12:17] can't reach two of our servers [12:19] amdprophet: doesn't appear to be, i'm on several VMs now [12:19] hmm maybe it's dns [12:30] amdprophet: figure it out? [12:37] up_the_irons: think the machine hosting our dns kernel panic'd [12:38] amdprophet: roger [12:46] tinydns makes it quite simple to have redundant dns servers :) [12:47] ie, you can just scp data and data.cdb around... [12:47] data in case you want to modify .. [12:48] mercutio: actually looking at using route 53 on amazon instead of hosting our own [12:49] up_the_irons: I can't get to the kvr10 vnc server [12:49] sounds scary to me amd :) [12:49] but each to their own [12:49] i wouldn't trust amazon with something critical like dns? [12:50] when not redundant.. [12:50] it is redundant [12:50] only across their network? [12:50] which means if they screw something up ? [12:51] maybe it's ok i dunno [12:51] amdprophet: each VM hosts its own VNC, so if you can't reach it, then you need to give it a hard reboot [12:51] it looks like anycast [12:51] up_the_irons: ack, I'll try giving it another hard reboot [12:52] wait, I see what the issue is now, I need to click boot after shutting down :) [12:52] yes :) [12:52] wasn't there a reboot button before? [12:52] i want to do my own anycast system heh [13:00] amdprophet: never had a reboot button [13:00] so yeh this direct vps peering thing up_the_irons ... [13:01] up_the_irons: my memory must just be going :P, everything is working now [13:01] does that mean i could advertise a /24 to you and you can advertise it out? [13:01] amdprophet: cool :) [13:01] mercutio: yes [13:01] would i also be able to tunnel traffic back to me? [13:02] i setup a gre tunnel last night as a little play, although then there is smaller mtu [13:03] although in my little checks it seems that 256mb is used on full route tables on 32 bit, 512mb on 64 bit. [13:04] mercutio: tunnel traffic back? [13:04] mm.. well i want to have some stuff terminate locally [13:05] and other stuff travel back to another location. [13:05] and advertise the same subnet in another loctation too [13:05] mercutio: you advertise a /24, it gets propagated to my peers / transits, that's it :) w/e you want to do beyond that is your choice [13:05] which also means tunneling the other wya [13:05] cool :) [13:06] when it said about 5 megabit commit or something [13:06] does that mean vps's can either have commited bandwidth or per gig charging? [13:06] that said.. i thought i might be doing heaps of traffic.. cos it say like 300 gig or something [13:06] then i realised that was over like a year [13:07] with like 1.5 megabit or something peaks over any of the diff measuring periods. [13:07] (i'm sure peak is higher than that on a shorter meausuring period) [13:08] i just figured out the serial console thing heh [13:08] it is actually kind of nicer being able to get in via ssh for oob over vnc [13:09] instead of vnc [13:10] mercutio: you _can_ have 95% billing with a vps, it's just that nobody understands 95% billing in the vps world and they prefer to see "per GB" pricing [13:10] ahh ok, 95% is what's showing atm? [13:10] in cacti? [13:11] mercutio: 95% and total GB are both shown on the graphs [13:11] yeh [13:11] i think total gb is cheaper for me atm :) [13:11] in general, 95% is cheaper [13:11] with low traffic volumes it noramlly is [13:11] hmm [13:12] 1.5 is charged at 2? [13:12] wut? [13:13] oh hangon [13:13] my 95th percentile is low [13:13] it's my burst that's high [13:13] my 95th percentile is 0.15 megabit :) [13:13] on monthly [13:14] and 0.8 megabit on daily [13:14] i mean you pay per megabit don't you? [13:14] so if you're 1.5 then you pay for 2? [13:15] if your commit was 1 mbps and 95th is 1.5, then you pay for 0.5 [13:15] bizzarely last about 8 hours ago there was a burst. [13:15] oh real? [13:15] so that'd be uhh [13:15] $30? [13:16] is it weekly graph i should look at for what it's worked out on? [13:16] it seems to show 4 weks [13:17] and that says 0.43 [13:17] graphs are always weird with hardly any traffic though [13:18] it's the monthly graph (we billing monthly) [13:18] but monthly seems to show year? [13:19] oh hangon [13:19] yeh you're right [13:19] it's above not below [13:20] monthly is 0.43 megabit [13:20] heh [13:20] i'm sure it'd go up just by having more ip's [13:20] with all the random scanning etc that happens now days [13:35] *** Webhostbudd_ has quit IRC (Quit: Leaving) [13:58] *** Webhostbudd has quit IRC (Quit: Leaving) [14:53] *** sako has joined #arpnetworks [16:05] up_the_irons: I see ads for arp all over the place now [16:06] pretty neat, it surprised me [16:41] congrats, you're eating his ad budget :P [16:42] you don't generally pay for impressions. you pay when they click. [17:01] don't you pay less if there are more impressions without click through? [17:27] *** sako has quit IRC (Ping timeout: 245 seconds) [17:56] *** heavysixer has joined #arpnetworks [17:56] *** ChanServ sets mode: +o heavysixer [18:42] mercutio: it's usually click thrus [18:43] *** dr_jkl_ is now known as dr_jkl [18:43] amdprophet: :) [18:45] *** Webhostbudd has joined #arpnetworks [18:51] up: yeh, but don't you pay less if there's a lower % of click through for the adword? [18:51] err less per impression [18:52] or is it just on click through direct [18:52] mercutio: just click thrus [18:53] ahh ok [18:53] i've never seen one of the ads myself, but i use adblock. [18:54] *** Webhostbudd_ has joined #arpnetworks [18:56] mercutio: you won't see anything with adblock :) [18:56] yeh, i don't really like random flash ads [18:56] or blinky things [18:57] animated gifs. [18:57] *** Webhostbudd has quit IRC (Ping timeout: 246 seconds) [19:38] *** zeeshoem has joined #arpnetworks [19:38] *** zeeshoem has quit IRC (Client Quit) [19:50] *** Webhostbudd_ has quit IRC (Quit: Leaving) [22:30] *** Ehtyar has joined #arpnetworks [23:06] *** zeshoem has quit IRC (Ping timeout: 248 seconds) [23:21] *** zeshoem has joined #arpnetworks