[01:15] an active channel is a happy channel [01:16] hehe [01:22] mercutio: urxvt +1 [01:23] phlux: nice screenshot [01:28] *** ziyourenxiang has joined #arpnetworks [01:39] up_the_irons: what kind of hw monitoring is on arp and what's on the client? [01:39] sorry - meant to say that this is regarding a dedicated server [01:40] ? [01:40] up_the_irons: well, if a disk dies for instance [01:40] ah [01:40] or a nic, or one of the two psu's [01:41] we have a nagios plugin for the disks [01:41] nic is not monitored (we encourage both nic's to be used in case one dies) [01:42] http://support.arpnetworks.com/kb/dedicated-servers/about-the-dual-1-gbps-gige-nics-on-arp-metal-dedicated-servers [01:42] that Source Code Pro font looks nice [01:43] so, all the client should care about is software? good to know [01:44] will most likely go to dedicated in june. What would a 128G ssd cost me/mo? [01:45] so the 128GB SSDs are the same price as 1TB SATA [01:45] ok, $20/mo then. sounds good [01:45] (haven't updated the website yet, but that will be the official pricing) [01:45] yeah [01:46] intel 7xx series? [01:46] depends on price, u have a link for one you like? [01:46] i was looking at these: http://www.newegg.com/Product/Product.aspx?Item=N82E16820227726 [01:47] damn, missed the sale... they were $89 a few days ago [01:47] i'm going to use mine for cache, so one of those msata things woudl work just as well for me. not sure there's supermicro boards for "enterprise" with those features just yet [01:49] (referring to these: http://www.newegg.com/Product/Product.aspx?Item=N82E16820167040) [01:50] well, doesn't matter much tbh. if you have something that works, lets go with that. [01:52] ok cool [01:53] what is the msata thing? :) [01:54] oh,, that link... [01:57] where the heck does that thing plug in? some PCI slot?? ;) [02:01] they have a special slot for them normally i think [02:02] cool [02:02] $89 seems insanely cheap for a ssd [02:02] i wondered if US pricing was diff than NZ pricing by that much [02:02] but newegg doesn't seem to have anything like that pricing normally, maybe they were getting rid of old stock [02:04] actually last i knew msata pricing was attractive for really small sizes, 20gb, 30gb etc [02:04] hmm corsair force is cheap [02:04] i have no idea what they're like [02:04] oh it's refurbished [02:05] $20/mo for a dedi ? [02:05] did I read that right [02:05] nesta: for a disk for a dedicated i think :) [02:05] o lol [02:05] * nesta goes back to observing [02:05] up_the_irons: ivy bridege has a designted thing for cache ssd's [02:05] heh ocz has more reburshied than other people [02:06] jbergstroem: ah [02:07] up_the_irons: talk about expanding to server boards as well, haven't seen it yet [02:07] jber: can't hotswap i imagine [02:07] they need little msata slots on the front or something [02:07] up_the_irons: called SRT or something. basically helps you to create a transparent drive on semi-hw raid plus one cache ssd [02:08] jber: only on windows isn't it? [02:08] mercutio: no, i think its hardware based. i tried booting os x on one which worked but i didn't really need it so i went for a single ssd instead [02:08] jber: oh, the help i meant [02:08] can't say i trust it 100% [02:09] it may passtrhough drive normally too [02:09] mercutio: ah, yeah probably. there's a lot of software/drivers built around it [02:09] it's evolving [02:09] mercutio: actually, os x has "fusion drive" nowadays too. something similar. http://rochetechnology.com/quick-hackintosh-tip-create-a-fusion-drive/ [02:09] i'd rather see flash-based-write-cache controllers bulit into motherboards [02:09] with passthrough disks rather than hw raid [02:10] mercutio: yeah. for sure. plus a small battery [02:10] or capacitor [02:10] the idea about having flash is that you only need enough charge to write the memory contents to flash [02:10] and then the battery doesn't have to last forever [02:12] jbergstroem: ah cool [02:13] *** ziyourenxiang has quit IRC (Quit: ziyourenxiang) [02:13] hmm amazon seem cheaper than newegg for ssd's [02:13] i was thinking of trying to buy one from the US before if it would be cheaper... [02:14] still more than $89 though [02:14] daym newegg for not shipping to aus/nz :( [02:14] they do [02:14] err [02:15] through third party ? [02:15] well you can get delivery addresses in the US [02:15] that then send to you [02:15] they repackage things now [02:15] to keep prices down [02:15] cos some places use huge boxes etc :/ [02:15] well nz has an easy to use service that isn't the cheapest [02:16] btu there are also international onces [02:16] it's still probably $10 to $20 for delivery even for something small like a ssd [02:17] url? [02:21] http://www.nzpost.co.nz/products-services/online-shopping/youshop?utm_source=home&utm_medium=midpage&utm_campaign=youshop [02:22] http://www.shipito.com/ [02:22] i've never used any of thse' [02:23] http://www.kiwishipping.co.nz/ [02:28] ok thanks [02:28] are you in au? [02:29] yeah [02:29] there's prob something local there [02:29] i assume demand is similar there [02:29] last time i compared, it basically added up comparing to local companies. preferred giving them my monnies instead [02:30] someone on local forum said shipito.com was expensive [02:30] you can easily pay > $200 USD for heavy things like servers [02:30] on various shipping things [02:31] but small things like ram/ssds/etc is where it probably could make more diff [02:31] although servers seem to cost a lot more in nz and au than the US [02:38] everything is more expensive :( [02:47] yeah [02:47] software is too :/ [02:47] even if downloading over the internet [02:48] btw, is it advised to run your own ntpd (i still do) even if you have a linux vm and run the kvm-clock tsc? [02:49] i guess it boils down to if the host runs ntpd or not? [02:56] i don't think running openntpd locally should be a problem [02:56] but i have no idea if it's necessary or not [02:57] you could compare running it versus not, but ram usage isn't that high [02:58] i compared at least 6 months ago, and localtime had drifted - so i installed openntpd (and nowadays run busybox-ntpd); but still, running ntpd if host already does is pointless. [02:58] (assuming you run kvm-clock) [02:59] what's kvm-clock? [02:59] its a driver for hte linux clock source [03:02] back in the days, people switched from the rc to a tsc from processors since it gave you a higher accuracy. kvm-clock is a way of getting hte host clock so the os doesn't have to keep track of this themselves [03:02] from the rtc, sorry [03:02] hmm i wonder what openbsd does [03:02] i still don't quite understand what hpet does [03:02] but it seems hpet is the normal timer these days for modern os's? [03:04] hpet is hardware timing, introduced in newer mobo's, replacing rtc [03:04] yep, its very reliable [03:06] but you can never avoid drift if you don't rely on ntp/clock protocols [03:07] pertty sure openbsd has support for both hpet and tsc [03:08] don't think it picks up kvm clock though. i just upgraded my 5.2 to 5.3 and still get unknown clock source [03:44] ahh [03:44] yeh old kvm version on most nodes [04:41] hmm [04:42] up_the_irons: you around? [06:08] *** cam13_ has quit IRC (Quit: Page closed) [06:30] *** heavysixer has joined #arpnetworks [06:30] *** ChanServ sets mode: +o heavysixer [06:45] *** heavysixer has quit IRC (Quit: heavysixer) [07:43] up_the_irons: To answer your question - mSATA stuff plugs into mini PCIe slots, but it's not strictly PCIe and the slot has to support mSATA. It's found inside most laptops nowadays, and I think I've seen it on some recent Intel desktop motherboards. Generally, mSATA is what laptops use for SSDs given the chance (Apple and Chromebooks to name a couple) since it's smaller and lighter and you could probably still fit an old-fashioned hard ... [07:43] ... drive too. [07:44] And I can vouch for the expense of shipping servers to NZ - My company's stuff is usually about 15lbs at shipping and costs the customer $300-$400USD for shipping (FedEx), plus duties. International shipping just sucks. [07:48] up_the_irons: btw I have a handful of those OCZ Agility3 120GB SSDs and so far they've all been fantastic! Been running them for about 18mos now and still going strong, and fast. Using them from everything - cache and log drives in ZFS pools, raid1 on my desktop (good lord the speed!), in the missus' gaming machine, and in most of my laptops. [07:48] Can't vouch for any of there more recent stuff (I know they changed controllers in the Agility 4, trading IOPS for throughput), but I'm happy. [10:02] *** heavysixer has joined #arpnetworks [10:02] *** ChanServ sets mode: +o heavysixer [10:10] *** nesta has quit IRC (Ping timeout: 245 seconds) [10:17] *** heavysixer has quit IRC (Quit: heavysixer) [11:53] *** hive-mind has quit IRC (Ping timeout: 256 seconds) [11:55] *** hive-mind has joined #arpnetworks [12:28] *** sean_ has joined #arpnetworks [12:29] if anyone uses openbsd… don't upgrade to 5.3 on arpnetwork vps' … you won't have a good time. [12:30] orly [12:30] What's the problem? [12:30] I use OpenBSD 5.3 on ArpNetworks and it's just fine? [12:31] Did you disable mpbios? [12:31] hmm… em0 just doesn't want to work. [12:31] watchdog timeouts everywhere [12:32] if i can try to intercept the boot loader i'm going to try mpbios. [12:33] I've had to disable mpbios since forever. [12:33] I don't know how you've been getting along without doing it. [12:33] (using console over ssh) [12:33] kvm page said it wasn't required anymore [12:34] this is maddening. :P [12:35] Do you still have an old kernel that you can boot? [12:35] CaZe: do you use the vnc console interface? doesn't seem very useful. [12:35] I use the console server. [12:36] I added a longer timeout to my boot script though. [12:38] * sean_ CaZe: how did you do that? [12:39] * CaZe sean_: echo "set timesout 60" >> /etc/boot.conf [12:39] *timeout [12:42] But I think I was able to still make it in at the default timeout, if you time it right. [12:43] it seems the console doesn't connect fast enough. [12:43] Maybe. [12:43] VNC should work though. [12:43] it swaps to com0 [12:43] (sry after) [12:44] set tty com0 [12:44] if i power off from a console session the vnc session disconnects and it doesn't reconnect fast enough at power on. [12:44] Can't you send a reboot through VNC? [12:44] just waiting now for config command to respond. [12:45] not sure. [12:45] Well, do you have an old kernel you can boot? [12:45] u using 32 bit or 64bit kernel? [12:45] maaaaybe. [12:46] 64. [12:46] I'm using an old snpashot though. [12:47] ah. just tried upgrade to release. [12:47] Well, try to get the timing right on the bootloader. [12:47] Or, just boot off the cd image. [12:48] (assuming you have an openbsd installer cd loaded) [12:48] Doesn't really matter which version installer. [12:48] yeah not sure. how can do you dhtat without access to the boo loader? [12:51] You have to get in on VNC. [12:51] k. [12:51] Just have your vnc client ready to login, with you login and password already typed in. [12:52] And when you boot your VM from the console server, wait like five seconds before clicking connect on VNC. [12:52] Maybe less. [12:52] Maybe not at all, I dunno. [12:52] Play around with it. [12:53] mpbios seemed to do the trick… lying KVM bastards [12:56] blargh. didn't fix icmp issue from 5.2 [12:56] What ICMP issue? [12:57] with pf enabled icmp drops, every 1 of 25 echo requests are returned. [12:57] (that's with an empty pf.conf) [12:57] disable pf.. and it works just fine. [13:01] one of these days i'm going to buy another vps to collect the info for a sendbug. [13:02] i've had this VPS since 4.7 so could be infrastructure related as well (ie. old KVM, old VM template etc.) [13:04] CaZe: thank or the boot.conf idea. definitely popped that in. [13:06] * sean_ CaZe: does your vps have the same pf behaviour with 5.3? [13:07] (using amd64 kernel) [13:07] (non mp) [13:07] What is it again? It drops pings? [13:07] ping 174.136.100.18 [13:09] Set pf debug logging… log fills with: /bsd: pf: icmp type 8 in wrong direction (1): ICMP out wire: [13:09] this happens from all over the internet so i'm pretty sure it's not just my devices. :P [13:10] Well, my snapshot is from August. [13:11] 100 packets transmitted, 5 packets received, 95.0% packet loss [13:11] * sean_ to me I assume? :) [13:11] Yes. [13:11] * sean_ it's just ICMP.. all other IP traffic is perfectly fine which makes sense. [13:50] *** scottschecter has quit IRC (Ping timeout: 245 seconds) [13:50] *** scottschecter has joined #arpnetworks [14:23] *** scottschecter has quit IRC (Ping timeout: 252 seconds) [14:25] who was complaining that i3 is 700kb yesterday? [14:25] because it's not true [14:25] ah, aslr [14:25] *** heavysixer has joined #arpnetworks [14:25] *** ChanServ sets mode: +o heavysixer [14:37] How is it not true? I du -sh'd the src directory, phlux. [14:37] *** scottschecter has joined #arpnetworks [14:38] maintainer claims it's mainly documentation that you're looking at [14:38] says the source is not 700kb [14:40] *** heavysixer has quit IRC (Quit: heavysixer) [15:42] *** scottschecter has quit IRC (Ping timeout: 256 seconds) [15:43] *** scottschecter has joined #arpnetworks [16:10] i3? ion3? [16:16] i3wm [16:16] .org [16:18] is it a fork of ion? [16:19] Doesn't look like it http://en.wikipedia.org/wiki/I3_(window_manager) [16:19] they prob should have used a more differnt name then [16:20] may be in tribute to it? *shrug* [16:21] There was a lot of excitement over I3 a few months back [16:21] hmm [16:21] i been using ion for over 10 years [16:21] so i suppose i'm not jumping at the chance to try other wm's [16:21] unless can see benefit [16:21] tried them all orig :/ [16:22] benefit: Not writing config in LUA [16:22] i don't do shit all lua, my config is simple [16:22] i don't even write a config, it writes it for me [16:22] but in the past i've changed some key bindings around [16:22] to execute different stuff [16:22] and that may or may not have involved lua [16:23] it is good to see more people getting into tiling window managers though [16:23] * brycec sticks to using Awesome [16:24] i found awesome meant constantly trying to rotate through different layouts [16:24] i prefer to just set layout myself [16:25] but i like they're using the new X stuff [16:25] Awesome isn't perfect :/ But I've made it work for me. Besides there's only a couple layouts I use consistently [16:27] Float, one of the tiling ones, and fullscreen [16:37] i am quite enjoying urxvt [16:38] only problem is i still haven't got the best colour scheme [16:38] it actually feels faster than gnome-terminal [16:38] and doesn't flicker [16:39] flicker? [16:39] xterm flickers [16:39] o [16:39] when you scroll [16:39] gnome-terminal flickers white when you make new terminal [16:39] urxvt does neither of those [16:39] roxterm also goes white when you make new terminal [16:40] (i use dark background) [16:40] URxvt*background: #000229 [16:40] have that as a background currently, it's like very dark blue [16:41] i suppose now my main issue is that ssh connectinos to remote hosts can be too slow to connect [16:41] i'm half tempted to use multiplexing which can speed it up, but can also mean they all die at once [16:42] it's partially cpu speed partially latency partially os [16:42] openbsd connects quicker than linux by about 2x for close hosts [16:42] O_o [16:43] 174 msce versus 279 msec, and the openbsd host is slower cpu [16:43] with faster connect [16:43] that's for ~10 msec away [16:43] it's 4.1 seconds to connect to uk host [16:43] #kiwiproblems [16:43] heh [16:44] yep [16:44] i dunno maybe something needs caching [16:44] DNS is the only thing you can cache here [16:44] i remember a while back there was a https optimisation that chrome started doing that reduced one rtt [16:44] but broke some sites [16:44] dns is anycast should be fast [16:45] it seems now days anycast is the main way to speed up dns [16:45] but hardly anything seems to do anycast on reverse lookups [16:45] and lots of dodgy sites have slow dns [16:45] like if you do lookups on random ip's that connect to you uninvited [16:45] that might be where its slowing down, openssh does reverse lookups on connecting IPs [16:46] yeah cache at remote end [16:46] but reverse dns isn't anycast [16:46] i wonder what ttl it has [16:47] 86400 [16:47] 2607:5300:60:e3a::1 - do a reverse on that, how long does that take you? [16:47] host 2607:5300:60:e3a::1 0.00s user 0.00s system 1% cpu 0.633 total [16:47] i'll try from somewehre else too [16:48] real0m1.795s [16:48] hmm [16:48] huge diff [16:48] both in nz [16:49] actually that may have been nameservers with 86400 ttl [16:49] i'm ahving problems finding the ttl using dig [16:49] that particular PTR has 1800 TTL [16:49] ahh [16:49] try timing a host lookup on 202.49.67.22 [16:49] *** HighJinx has quit IRC () [16:50] host 2607:5300:60:e3a::1 0.00s user 0.00s system 1% cpu 0.322 total [16:50] ;; Query time: 633 msec - resolving via local recursor [16:50] host 202.49.67.22 0.00s user 0.00s system 1% cpu 0.462 total [16:50] curious [16:50] bryce has faster dns :) [16:50] :D [16:50] that's actually interesting [16:50] static got same query time for my reverse lookup, as i got for his reverse lookup [16:51] (dnsmasq on my router, spread between Google's public DNS servers v4 and v6, as well has HE's DNS servers v4 and v6) [16:51] ;; Query time: 0 msec - now its in my cache :) [16:51] bryce: using all-servers? [16:51] lol staticsafe [16:51] I run my own DNS recursors [16:51] ahh i used to do that [16:51] because why not :) [16:51] using unbound? [16:51] BIND because i do authoritative as well [16:52] mercutio: yes, all-servers [16:52] eww :/ [16:52] brycec: that's what i shifted to doing at home [16:52] *shrug* I've taken a liking to BIND [16:53] i wonder why that other server i tried took 1.7 seconds [16:53] probably didn't have cache of some steps before hand [16:53] yep [16:53] cold caches [16:53] that's one of the reasons i don't run my own dns resolver [16:54] I find that pointing my home network towards my 2 resolvers builds up a nice cache :) [16:54] i wonder how many qps before it makes sense? 20? 100? 200 ? [16:54] heh [16:54] i benchmarked heaps of dns resolvers before [16:54] then found that the way i was benchmarking didn't really measure real world performance [16:54] it's a kind of complex issue [16:54] http://stats.asininetech.com/asininetech.com/uriel.asininetech.com/bind9.html - this one is also my mail server, so lots of queries [16:55] like no-one in nz probably looks up domains in ethiopia [16:55] so everyone will have slow lookups to there [16:55] yes [16:55] but that doesn't matter [16:55] so if you randomly pick domain names [16:55] you'll probably find some international focused dns stuff will be faster than local stuff [16:55] but if you pick real usage domains, closer will be better [16:55] like facebook has a high chance of being cached [16:56] I originally started doing my own DNS because my ISP's one was so shit [16:56] do you run web stats? [16:56] i see lots of spikes [16:56] heh [16:56] web stats? [16:57] like looking up heaps of reverses [16:57] using webalyser or such [16:57] http://mx1.stats.staticsafe.ca/ [16:57] in a cron job [16:57] nope [16:57] hmm [16:57] i wonder what the spikes are then [16:57] more e-mail [16:57] ahh pflogsumm [16:57] wow you hardly have any rejections [16:57] my rejection rather is like 10x my received or something [16:58] s/rather/rate/ [16:59] heh what'st htat SASL LOGIn thing [16:59] someone trying to brute force you? [16:59] yea noticed [16:59] not unusual [16:59] oh actually my rejection rate went down a lot [16:59] still higher than yours [17:00] as you can see, its my mailing list e-mail, so not a lot of spam [17:00] date received delivered deferred bounced rejected [17:00] -------------------------------------------------------------------- [17:00] Apr 28 2013 957 675 109 7 9105 [17:00] Apr 29 2013 1611 1116 199 20 5606 [17:01] Apr 30 2013 3762 2557 175 1 1580 [17:01] actually 0 spam from non-list sources [17:01] it's kind of ireregular it seems [17:02] my reject count is unusual, its actually just greylisting [17:02] oh [17:02] i use zen.spamhaus.org [17:02] which im thinking of disabling [17:02] i don't greylist anymore [17:02] it's too annoying when you're waiting for a mail [17:02] yea i use postscreen with zen.spamhaus.org and some other black lists [17:02] oh so it just doesn't reach postfix [17:02] yea [17:03] postscreen is <3 [17:03] but yeah, i dunno why apr 28 was so bad [17:03] and apr30 good [17:03] it won't do much harm, it's just one of those things i occassionally notice [17:03] i imagine peoelpe without zen.spamhaus.org get a lot more spam? [17:04] yea [17:05] zen kills a lot of stuff [17:05] it's annoying how wasteful of resources mailing lists are [17:05] but as ssd's get cheaper etc i suppose people are likely to care less and less [17:05] amavisd-new takes care of the rest [17:05] i'm using amavis too [17:05] and dkimproxy [17:06] I use opendkim for DKIM checking and signing [17:06] and dcc, and razor, and pyzor etc. [17:06] i used to use dspam [17:06] I was thinking of dspam but meh [17:06] i'm actually still getting some phishing coming through [17:06] i dunno dspam started going wrong [17:06] and amavis can't feed it anymore [17:07] it expires tokens [17:07] so if you stop feeding it stuff it goes weird [17:07] i think that was it [17:07] yea possibly [17:07] it's obvious phishing stuff though [17:07] i dunno how to block it easily though [17:08] 57444 N May 04 Kiwibank ( 254) Kiwibank internet banking customer support [17:08] 51094 O Mar 11 ASB Bank ( 73) Hello ASB Customer [17:08] 57338 O May 02 ASB Bank Limite ( 132) Irregular Activity On Your Account. [17:08] etc [17:09] i actually surprisingly little spam on even my main accounts [17:09] get* [17:09] bank phishing seems popular [17:09] and fedex [17:09] and paypal [17:09] i'm not even using kiwibank or asb bank though [17:10] the real problem is that lots of bank email looks like spam too [17:10] with html and crap [17:10] heh [17:10] sent from weird remote mail servers [17:10] my bank sends me 1 e-mail per month [17:10] hmm so does mine [17:10] saying i should pay my credit card [17:11]