=== rww is now known as rwd === rwd is now known as rww [00:21] sarnold, thank you. I've run into a similar info wall googling [00:21] nacc, physical, but as far as pg is concerned, both. Though I'm told logical cores don't always work out as well as expected. [00:21] jpastore: *nod* not a real surprise, the folks who know just do what they need to do... [00:22] sarnold, /sigh so how do those who don't know become the ones who know? [00:23] it really depends on the workload [00:23] jpastore: if you're lucky you find the right group to help you along :) [00:23] ht cores only help, when your using different parts of the cpu [00:24] patdk-lap: so, there's a question -- which would you prefer of two e5-26.. v3 CPUs, one with ten cores and no HT, or eight cores and HT? :) [00:24] those are small options [00:24] patdk-lap, well, the application is telephony switch. I'm setting kamailio for LCR. so every dial is going to be a few queries. [00:24] need to know clock rate, interchange speed [00:24] qpi speed [00:25] and specifically, what load it will be running [00:25] I'm storing CDRs, querying for LRN to feed LCR. [00:25] some loads do just math, so more cores better [00:25] orthers need more memory access, so qpi matters [00:26] well this would be a few read do hi/lo btree comparison in the index, and then a write for the cdr for later analytics [00:26] if you want max lookup speeds [00:26] and lookups won't hit in the cpu cache [00:26] you want best qpi speed [00:26] fastest access to memory [00:26] ok I'll check on the chips quoted [00:27] if the issue is running many many at once [00:27] it might be more cores is better [00:27] then the debate is cpu cache vs memory acccess speeds [00:28] memory access is slow, and that core will wait, ht could help make use of that core during that waiting though [00:33] sarnold, if all things the same, 8cores [00:33] ht is hardly benifit pg itself [00:33] it will help the system some, but not enough vs 2 more real cores [00:34] ht at the best normally gives 20% improvement, [00:34] patdk-lap: so, here's the two systems.. http://www.cpu-world.com/Compare_CPUs/Intel_CM8064401612900,Intel_CM8064401831000/ [00:34] oh ya, 10core :) [00:35] patdk-lap: 2630 vs 2663; the 2663 chips look like they're impressive beasts but I don't know if they'd be faster enough for my work loads to justify the extra heat.. [00:35] higher qpi, but only alittle [00:35] much higher cache [00:35] the qpi you can see translates to supported memory [00:35] ddr-1600 vs ddr-2133 [00:36] some years back, ht was a joke. the latest tests I've seen show it may be a big gain, obviously depending on type of load [00:36] so if we give ht it's best case of 20% improvement, that is like 1.6 extra cores at best case [00:36] I -think- that the systems that are using the 2663 have all 24 memory slots occupied and thus run at slower speeds anyway [00:36] sarnold, only if you fill them [00:37] if you don't fill every slot, it will run full speed [00:37] RoyK: yeah, it used to be something like 5% performance boost or 10% performance loss.. but these days with super-deep pipelines and glacial memory speeds, it might make more sense than it used to [00:37] think it's more that the compilers are much more tuned to making ht friendly code too [00:38] removing as many jumps and stuff as possible, so the pipeline doesn'tget dumped and reloaded [00:38] a friend of mine tested his i7 something for transcoding video and he got a good boost by enabling ht (after we were discussing how much it really can do) [00:38] patdk-lap: the nice thing about filling all the slots is that it's a hueg amount of memory :) http://www.ebay.com/itm/IBM-LENOVO-THINKSERVER-RD650-2x-E5-2663-v3-10-CORE-192GB-12x-3TB-SAS-W-RAILS-/201468817323?hash=item2ee87a2fab [00:38] I beleive that boost was 40% or so over non-ht [00:38] RoyK: wow [00:39] sarnold: I don't remember the numbers correctly - posted a question to him - guess he'll be up tomorrow sometime to answer it [00:40] RoyK: thanks [00:46] sarnold: I looked at the IRC logs from back in may - looks like it was some rendering with a performance gain of 35% or so. Well hopefully know more tomorrow [00:46] RoyK: wow. very impressive :) [00:47] it surprised me a great deal [00:48] it's certainly a lot more than I'd expect. but I don't know how well rendering matches my own eventual needs.. VMs, fuzzing, ubuntu archives, and all the packages unpacked for archive-wide greps, cppcheck, shellcheck, etc runs... all those feel more disk-blocked than RAM blocked [00:48] but even then, another hardware thread willing to run might benefit greatly from the HT [00:52] I just don't know enough about this... [00:52] ya depends, those workloads are so random [00:52] if it was just postgres, I would go the 10core [00:52] single threaded performance, faster memory speeds, would greatly help it [00:53] I went to an HPC seminar some 4 or 5 years back and they were like "turn it off!" [00:53] for postgres, I'd leave it on [00:53] ya, I would always leave it on [00:53] but the question is 8core with ht, or 10core without ht support :) [00:53] 2.4ghz vs 2.8ghz [00:54] although... I don't have much postgres servers left that aren't virtual ;) [00:55] We have some 250 servers in operation, some 30 of them are physical and most of those are to be retired soon [00:59] RoyK: heh, are any of them going to be for sale? :) [01:01] oh, I forgot, you're in norway, right? shipping/import probably too much hassle :) [01:12] sarnold: mostly old stuff - the newest ones we're throwing out will be R300s [01:13] all new servers are blade servers anyway, except a handful of 1U or 2U servers (and one 4U thingie with zfs for cctv storage) [01:21] RoyK: ahhh :) [01:23] some of our old school admins still stick to physical servers (even though those are the most troublesome ones) [03:14] anybody know how to set up... libaprutil1-dbd-pgsql [03:14] I cant figure out where the config file is... [03:16] does dpkg -L libaprutil1-dbd-pgsql help? [03:28] let me check [03:38] ya no good [03:38] i just dont understand [06:00] pmatulis: sorry, was out today. i don't know of any iptable sissues with using lxd. lxd is afaik still network-agnostic atm. It will eentually have a ipv6-only bridge of its own, but not yet. [06:00] pmatulis: waht are you seeing? [08:21] hello ,I want to put the server.iso file inside the Ubuntu server throught command how can I do that please "the ubuntu server is running on VM and the OS is windows" I want to put it in dirctory /var/lib/libvirt/images??? [08:31] I would like to copy file server.iso to dirctory /var/lib/libvirt/images ?? I have a file in usb flash disk and the OS is windows [08:43] haidar: to clarify, you have a flash drive plugged into your linux server, and you want to copy a Windows Image from that to /var/lib/libvirt/images? [09:25] Hi [09:26] I'm getting GPG errors when doing apt-get update [09:26] http://dpaste.com/1JT6GST [09:56] hi all [09:57] i configure my main board for use raid 1 with two hard driver and install ubuntu 14.04 server and in boot i get these error [09:58] mdm create use root not found [09:58] mdadm create groud disk not found [09:59] and system not boot and these message print in a loop === cpaelzer is now known as cpaelzer_afk === cpaelzer_afk is now known as cpaelzer [12:09] hallyn: i'm not seeing anything yet. i'm writing LXD:Juju documentation and i'm trying to be proactive === saldkjas is now known as utis === utis is now known as ucionas === coreycb` is now known as coreycb [14:53] jamespage: would the plan be to backport 0.80.11 to the precise Cloud archive (unlike 0.80.10 which didn't make it?) LP#1535278 [14:54] gQuigs, for 12.04? [14:54] bug 1535278 [14:54] jamespage: yup [14:54] bug 1535278 in ceph (Ubuntu Trusty) "0.80.11 stable point release" [Medium,In progress] https://launchpad.net/bugs/1535278 [14:55] gQuigs, once SRU team accept that into trusty, I'll get it pushed through to icehouse-proposed for UCA [14:55] gQuigs, we'll give everything a test and then push out that and any other updates that are pending release [14:56] jamespage: awesome :) [14:56] I was just concerned because 0.8.10 never made it out of the cloud archive -proposed it seems [14:56] .10 not being released was an oversight - apologies [14:56] oh ok :) [14:56] thanks! [16:54] pmatulis: oh. well today/tomorrow stgraber and tycho and i will be traveling, but we'll be all in one spot thu/fri so might be a good time to have a hangout to discuss docs [16:54] hallyn: ok [16:59] stgraber: do you care when/what time we would do that? pmatulis: you're east coast? [16:59] hallyn: yeah [17:01] pmatulis: ok i'll schedule something for thu - morning your time. [17:01] hallyn: are there other things you want to discuss re juju/lxd docs? lxd/firewall may be just a few minutes [17:03] pmatulis: do you have a rough draft you can send us so we can think about it? 'just a few minutes' is definately fine :) But i wanna make sure we have kickass lxd doand juju+lxd docs on help.ubuntu.com [17:04] pmatulis: invite sent - thanks for working on the docs! === Lcawte|Away is now known as Lcawte === cpaelzer is now known as cpaelzer_afk [17:56] hallyn: lxd itself will remain on linuxcontainers.org for now and juju/lxd docs will be on jujucharms.com (that url always feels strange) [17:56] hallyn: maybe you can review the juju/lxd stuff - https://jujucharms.com/docs/devel/config-LXD [18:10] pmatulis: got it open in a browser, will look over it at airport later this eve - thx [18:11] hallyn: thanks [18:15] w 23 [18:15] excuse me. === cpaelzer_afk is now known as cpaelzer [18:51] is there a reason why tomcat8 isn't available from the default repos? [18:52] DammitJim: it's in universe? [18:52] DammitJim: what do you mean by "default"? [18:52] like with the normal repos [18:53] like in the sources.list [18:53] oh [18:53] I'm on trusty, sorry... talking LTS [18:54] DammitJim: ah, one sec [18:54] I'm only asking because everyone at my company keeps saying we should be on the latest [18:54] well, I don't want to go to the latest if there is a reason why it isn't in the normal repos [18:55] DammitJim: tomcat 8 was only uploaded to debian unstable a few weeks before trusty was released - way too late to make it in [18:56] DammitJim: yeah, looks like it wasn't added to ubuntu until vivid, so it wouldn't be in that LTS [19:05] oh ok [19:06] but it's not like it's not "trusted" or "stable" software [19:06] I just have to add the proper repo, right? [19:06] um, what? [19:06] DammitJim: it's not available in Trusty. Unless there's a third party repository providing the software, it won't be available in Trusty [19:06] and you'd have to update to Vivid if you want the copy that's there [19:06] or rather Wily, 'cause I think Vivid is EOLing soon [19:07] I would be upgrading to the next LTS release [19:07] then you wait for 16.04 [19:07] then upgrade to that [19:07] is that coming soon? [19:07] in 2016/04, as the version implies :P [19:08] duh for me [19:08] I never knew that [19:08] look at that! [19:08] heh I was a bit embarassed when I finally figured it out too :) [19:09] DammitJim: there is also a process for requesting backports https://wiki.ubuntu.com/UbuntuBackports and a tool for doing so - requestbackport(1) from ubuntu-dev-tools [19:10] any sys admins in da house? [19:10] DammitJim: there is also backportpackage(1) in ubuntu-dev-tools, for building a backport locally or in a ppa [19:10] I think I'm going to go nuts!!! [19:10] thanks for the info tarpman [19:10] I'm in a weird dilema [19:10] as a company apparently, we need to strive to be on the latest release if possible of any software we run [19:11] I work in the IT department and the developers are telling me that I'm the one who needs to determine what version of tomcat should be installed on the servers [19:11] what is wrong with that last statement? [19:13] * tarpman shrugs [19:13] you're running the latest (LTS) version of ubuntu [19:13] and software provided by ubuntu and supported by its security team [19:13] if you diverge from that, you're taking on all the burden of tracking security updates and things yourself [19:14] it's not hard for even a manager to understand ;) [19:14] fwiw it looks like tomcat8 is still in universe in xenial anyway [19:15] ah yeah, was looking at tomcat7 - still getting some security updates in trusty and vivid [19:15] what does that mean sarnold ? [19:16] DammitJim: packages in universe get security support from the community -- there's a handful of packages where someone from the community cares enough to keep it up to date, but most don't get that kind of attention [19:16] DammitJim: https://help.ubuntu.com/community/Repositories/Ubuntu [19:16] oh, so that's a good thing, right? [19:17] thanks nacc [19:17] so, in terms of determining the version of tomcat to use, do you guys think that's a sys admin thing or a developer thing? [19:17] or together? [19:17] DammitJim: well, it surely beats having -no- repository of software :) but unless someone steps up to provide e.g. tomcat8 patches, it just doesn't get updated. [19:17] I hate the fact that I was thrown the words: "We don't care as long as it is tomcat, you figure out the version" [19:18] which is crazy because they are the ones who build the software that has to be compatible with tomcat [19:18] or am I totally wrong? [19:19] DammitJim: they may stick with a core of functionality that they expect to work in tomcat7 and tomcat8 alike; or they may be willing to tailor the software to fit the requiements of either one. You need to figure out the best mixture of upstream support, ubuntu support, and self-provided support; it's not necessarily easy [19:22] dammit!!! [19:22] ... jim [19:24] so, it is up to me, not the developers? [19:25] it might be worth a discussion :) [19:25] ok, good [19:25] I think I'm going to collect all the information you have given me [19:25] so, release schedule [19:25] support [19:26] "hey guys we can use canonical-supported tomcat7 packages in the latest LTS will very little effort" [19:26] upstream plan [19:26] anything else I"m missing? [19:26] vs if I need support for tomcat8, where would I go? [19:26] "but it'd be a lot of effort for me to maintain tomcat8, if that's somethin gyou actually need. do you need any of its features? is it worth the cost?" etc [19:26] in terms of security... tomcat 7 and tomcat8 are probably the same, right? [19:27] probably close enough [20:14] My question is regarding package guidance of samba, not support per se. [20:16] Asking here because ubuntu-server is the only team responsible for that package that has a channel. [20:16] I noticed Xenial has been bumped up to 4.3. Yay! [20:17] 4.4 will be released in March. [20:17] dcnoderunner: what's your question? [20:18] Samba does not do LTS type releases. To get the longest support from them upstream, one should pick the "freshest" release. [20:18] Has any finite decision been made on if Xenial will ship with 4.4? [20:20] dcnoderunner: feature freeze for xenial is feb. 18, just fyi [20:20] (and where do I find such info in the future so I don't have to bug you guys each release?) [20:22] dcnoderunner: https://wiki.ubuntu.com/KernelTeam/Newsletter/2016-01-12 [20:22] FYI received. [In the back of my head I think I've seen things arbitrarily exempted by Canonical employees in the past] [20:23] Yay again! [20:23] dcnoderunner: https://wiki.ubuntu.com/Releases usually has links off to per-release plans, schedules.. [20:23] but the wiki seems unhappy ATM. [20:24] Oh wait, that meant kernel 4.4. Oh well, that's good too. [20:34] So the firm-but-not-final answer I'm hearing is wait and see what's been pulled or not by Feb 18th. [20:35] I'll optimistically stay in channel to hear more for a few more hours.. [20:38] dcnoderunner: samba 4.3.3 is what's in xenial right now [20:38] if 4.4 isn't GA until March, I'd expect it won't make Xenial [20:40] Yeah, and I'm happy you're at least willing to break with Debian to bump it up to that. [20:41] dcnoderunner: to be clear, it's no break ... current+1 follows unstable until feature freeze (when also, I believe the debian autosync is turned off) [20:41] dcnoderunner: and unstable (sid) now has 4.3.3 as well [20:42] But samba only supports their releases for 18 months. So the longer you're 'in the range' of their support cycle, the easier it should be to work with them and squash bugs. [20:45] https://wiki.samba.org/index.php/Samba_Release_Planning [20:45] https://wiki.samba.org/index.php/Release_Planning_for_Samba_4.4 [20:48] dcnoderunner: there's lots of unsupported (upstream) versions potentially -- i would consider send an e-mail the -devel mailing list to find out, i'm not sure who's in charge of deciding that [20:48] nacc: Oh, really? Unstable was still at 4.1 when I looked last week. Sorry. [20:52] dcnoderunner: np, just looked today, so don't know when it got updated [20:56] https://tracker.debian.org/news/734141 says samba 4.3.3 was accepted into unstable back in december... [20:57] tarpman: thanks [20:58] dcnoderunner: fwiw, distro support for a particular package doesn't necessarily end when upstream support does. distro maintainers tend to have lots of experience supporting their stable releases by backporting patches and such. [20:59] rbasak: let's say we did decided to have both php5 and php7 available; how would we deal with the pear/pecl packages? would there then be two versions of each? or would we simply not ship php7 versions? [22:26] nacc: I imagine we wouldn't ship php7 versions [22:26] nacc: I think it would be even more too much work to try and supply two sets of each. [22:30] rbasak: yeah, that's what i figured -- it would get pretty hairy, i think