| === thumper-gym is now known as thumper | ||
| === jaimef_ is now known as jaimef | ||
| === jaimef is now known as Jaimef | ||
| === Jaimef is now known as jaimef | ||
| MavKen | is it ok to regularly run apt-get dist-upgrade? | 01:27 |
|---|---|---|
| sarnold | MavKen: see the unattended-upgrades package :) | 01:28 |
| MavKen | ok | 01:28 |
| Bilge | Why, when installing a package for which configuration conflicts with a file already existing on the system, do you only get prompted to explore the differences the first time the package is installed? Subsequent remove/installs do not prompt for conflicting configurations | 01:29 |
| sarnold | Bilge: it depends if the packaged configuration file changes | 01:33 |
| sarnold | Bilge: if the file doesn't change on a package upgrade, it's left alone. and since we all dislike those prompts, most packages try to avoid changing those files on a whim :) | 01:34 |
| Bilge | Right, but I just installed a package and got the prompt | 01:35 |
| Bilge | Then I removed it and installed it again | 01:35 |
| Bilge | No prompt | 01:35 |
| Bilge | Why? | 01:35 |
| Bilge | I want to get the prompt back | 01:36 |
| sarnold | Bilge: try this: echo >> /path/to/config/file -- changing your local copy of it (hopefully appending a newline doesn't change the meaning of the file) might bring the prompt back | 01:37 |
| sarnold | it might only prompt again if the upstream version changes though. I can't recall the details. | 01:37 |
| Bilge | touch isn't enough?> | 01:41 |
| sarnold | that won't change a cryptographic hash of the file | 01:42 |
| Bilge | It didn't do anything anyway | 01:42 |
| Bilge | Normally it is only when the upstream changes but there MUST BE a setting locally that flags whether the prompt has been shown | 01:43 |
| sarnold | Bilge: try reinstalling the package with --force-confask -- see the dpkg(1) manpage for details. I'm not convinced it'll do the job but it's worth a try | 01:45 |
| Bilge | I install the package with apt-get, and it does not understand that switch | 02:13 |
| Bilge | I don't know how to install with dpkg | 02:13 |
| Bilge | It wouldn't help anyway since I want to reset the prompt for another "user" (a robot), not for my use | 02:14 |
| sarnold | Bilge: there'll be a package in /var/cache/apt/archives/ for the package you're installing; you can use dpkg -o on that | 02:15 |
| sarnold | Bilge: hrm, how about make a copy of the file; install the package; apt-get purge the package; move the backup back in place; try the install again | 02:15 |
| Bilge | Purge did the trick :) | 02:19 |
| sarnold | yay! | 02:19 |
| sarnold | I'm sorry I didn't think of it earlier. sigh. | 02:19 |
| sarnold | give me an hour, I'll eventually come around to the right answer. sheesh. :) | 02:20 |
| Bilge | I appreciate the help | 02:22 |
| Bilge | Can't beleive I didn't know about the purge command either | 02:33 |
| === justizin_ is now known as justizin | ||
| === thumper is now known as thumper-back-for | ||
| === thumper-back-for is now known as thumper-afk | ||
| Voidal | hey guys, odd question but is it possible for apache to cache files on the server? (Sorry I'm a bit new) | 06:30 |
| Voidal | my university assignment would display data from hours ago at random | 06:30 |
| sarnold | there's a lot to modern web contraptions... it's common to run a caching proxy in front of web servers to spread the load across more machines.. | 06:32 |
| TJ- | Voidal: Depends on what the Expires header says | 06:32 |
| TJ- | Voidal: See mod_expires | 06:32 |
| Voidal | I'm not sure we students have access to that, my lecturer said the MIME type on their server might be not expiring | 06:33 |
| Voidal | It's all submitted and working now, it's just been bugging me | 06:33 |
| Voidal | sarnold: that might be it, it's a large university with a ton of computers | 06:34 |
| sarnold | if the mime types had something to do with it, I'd suspect a broken proxy -- caching decisions should be made on the basis of more concrete things than what "usually" works for a given mime type.. | 06:34 |
| sarnold | and who knows how the web server gets the files it serves; if they are loaded over nfs, and it caches nfs aggressively (against spec but common because nfs would otherwise be slow :) -- it might be working fine but a lower level might not have been working quite right | 06:35 |
| TJ- | Yeah, if the user-agent does a HEAD and the there is no Expires itwon't bother fetching the document, it'll use the one in its local cache | 06:36 |
| TJ- | You can usually force the browser to do an explicit refresh (GET) in that case, I seem to recall Ctrl+F5 on some browsers would do it | 06:37 |
| Voidal | Well shit, I wish I knew about ctrl+f5 yesterday | 06:37 |
| Voidal | I enjoy php programming a bunch but my experience with servers is only basic | 06:38 |
| Voidal | so this is all pretty new :) | 06:38 |
| TJ- | For dynamically generated content (e.g. PHP scripts), if the result is being cached then generally the server is mis-configured | 06:38 |
| sarnold | well... | 06:38 |
| sarnold | generating generated content is also pretty slow | 06:39 |
| sarnold | there's a lot of opportunities for caching content there, as well, and most frameworks will have mechansisms to make that happen | 06:39 |
| Voidal | the assignment generated response data (in xml and json) by pulling data from a source XML file | 06:40 |
| sarnold | fragment caching, full-page caching, etc. knowing when to expire that cached content can be a huge part of keeping such a system afloat and functional | 06:40 |
| sarnold | Voidal: for more than you'll want to know :) check out section 14.9 here: http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html | 06:41 |
| sarnold | time for bed, have fun Voidal and TJ- :) | 06:41 |
| Voidal | thanks for the help | 06:42 |
| MangledBlue | Can anybody help??? simple install - my MD5 checks out in ISO- c7f439e864d28d9e5ca2aa885c4ec4cb *ubuntu-12.04.4-desktop-amd64.iso - but I cannot get it to work - any thoughts? | 06:59 |
| === _ruben_ is now known as _ruben | ||
| === rvba` is now known as rvba | ||
| === LeMike is now known as LeCappu | ||
| rcsheets | I'd like to start testing PXE auto install for 14.04, but it appears that none of those bits are present on the current daily build CD images. Does anyone know when that might be expected to change? | 07:50 |
| Lord_Set | rcsheets, why not use MAAS? | 08:08 |
| Lord_Set | Any developers here that deal with network drivers? | 08:11 |
| Lord_Set | I have a question regarding a specific 10g NIC/HBA/CNA | 08:12 |
| rcsheets | Lord_Set: I have not investigated MAAS, because the name makes me think it is for deploying physical systems. I am currently focused on deploying VMs. | 08:16 |
| Lord_Set | Ahh ok. MAAS will work for deploying VMs as well as far as I know. | 08:17 |
| rcsheets | Lord_Set: I would prefer to tweak my existing, working VM deployment workflow so that it works with 14.04, rather than switching to something entirely new. But I can change if support for PXE/casper has been dropped. | 08:17 |
| Lord_Set | MAAS/Juju should be able to deploy everything you need... | 08:18 |
| rcsheets | alright well if I have to set up entirely new stuff, then I guess I'll give up for now and wait till I have more time. | 08:22 |
| TJ- | rcj: what bits seem to be missing? | 08:22 |
| rcsheets | TJ-: was that for me? | 08:22 |
| TJ- | rcsheets: oops! bad tab-completion! what bits seem to be missing? | 08:23 |
| rcsheets | TJ-: /install/netboot doesn't exist. that's where i normally find a kernel and initrd suitable for PXE-booting the install environment | 08:23 |
| TJ- | Hmmm, I have a script the breaks out a standard installer ISO for PXE, not noticed any issues with Trusty ... let me look | 08:26 |
| rcsheets | TJ-: maybe I'm just not seeing where to get the alternate install iso? all i can find is desktop. | 08:28 |
| TJ- | There is no alternate any longer | 08:29 |
| rcsheets | server and alternate are the only ones i've seen pxe bits on, and i don't see daily isos for either of them | 08:30 |
| rcsheets | oh! trusty server is at http://cdimage.ubuntu.com/ubuntu-server/daily/current/ | 08:30 |
| rcsheets | well then, i got the wrong iso. silly me | 08:30 |
| * rcsheets looks embarrassed | 08:30 | |
| rcsheets | yep, on the proper iso everything is as i expected it to be | 08:34 |
| rcsheets | sigh | 08:34 |
| === thumper-afk is now known as thumper | ||
| === psivaa-afk is now known as psivaa | ||
| nitin | hello | 11:05 |
| nitin | hello .. guys is it worth to configure a vps running ubuntu server where apache is actually running and a website is being hosted to be configured as a VPN server as well..? | 11:06 |
| leitmedium | nitin: what do you mean by "is it worth"? technically it is possible, of course. A simple well-documented OpenVPN setup is enough. | 11:40 |
| leitmedium | nitin: if you want to use the server encrypted, you can go with ssh and https, too. | 11:40 |
| hello123 | hello | 11:48 |
| hello123 | i wanted help in software raid | 11:48 |
| hello123 | any 1? | 11:49 |
| hello123 | ? | 11:51 |
| leitmedium | hello123: it's better to just type your question. | 11:55 |
| hello123 | ok | 11:55 |
| hello123 | i have 6 x 4tb drives | 11:55 |
| hello123 | but in/home 4 tb is showed | 11:55 |
| leitmedium | what das "cat /proc/mdstat" say? | 11:56 |
| leitmedium | please paste to paste service :) | 11:56 |
| hello123 | http://pastebin.com/Mb0CLYQp | 11:57 |
| hello123 | :) | 11:57 |
| leitmedium | is it possible you setup a raid1? | 11:58 |
| leitmedium | just mirrored? | 11:58 |
| hello123 | i didnt setuped | 11:58 |
| hello123 | while installin os it automatically came | 11:58 |
| hello123 | i need 24tb @ /home | 11:58 |
| hello123 | :( | 11:59 |
| leitmedium | hello123: check: "dmesg | grep md2" and "dmesg | grep md3" | 12:00 |
| leitmedium | there you shoud see somethink like | 12:00 |
| leitmedium | "md/raid1:md2: active with 2 out of 2 mirrors" | 12:01 |
| hello123 | letitmediuim i messed with partition and now reinstalling os | 12:01 |
| leitmedium | this gives you a hint, which type of raid you have | 12:01 |
| hello123 | just in less than half an hour it will be back | 12:01 |
| hello123 | its 3 rd time reinstallinjg | 12:01 |
| ikonia | /proc/mdstat will show you each arrays status | 12:01 |
| ikonia | as will mdadm exaine | 12:02 |
| ikonia | examine | 12:02 |
| leitmedium | ikonia: but not the type, right? | 12:02 |
| ikonia | the type ? | 12:02 |
| ikonia | that will show you the raid level | 12:02 |
| ikonia | or should do | 12:02 |
| hello123 | i have sent it | 12:02 |
| hello123 | check in pastebin | 12:02 |
| ikonia | which one ? | 12:02 |
| ikonia | md5 : active raid1 sda7[0] sdb7[1] | 12:02 |
| ikonia | (as an example | 12:03 |
| hello123 | http://pastebin.com/Mb0CLYQp | 12:03 |
| ikonia | shows my md5 is a raid 1 mirror made up of 2 disks on partition 7 | 12:03 |
| leitmedium | ikonia: of course, just overlooked it | 12:03 |
| leitmedium | ikonia: you are right | 12:03 |
| hello123 | first pic | 12:03 |
| ikonia | hello123: yours are both mirrors | 12:03 |
| ikonia | they are all raid 1 | 12:03 |
| leitmedium | hello123: you have a raid1. so every bit of data is replicated throughout all disks | 12:03 |
| hello123 | @leitmedium yahoo or skype? | 12:04 |
| hello123 | yeah | 12:04 |
| hello123 | i want it raid 5 | 12:04 |
| leitmedium | hello123: at least this is very fault tolerant | 12:04 |
| leitmedium | :) | 12:04 |
| hello123 | and full 24tb @ /home | 12:04 |
| leitmedium | hello123: then you have to resetup the array | 12:04 |
| hello123 | i dont know thats the problem new to linux :( | 12:04 |
| hello123 | tried and it has been almost a month | 12:05 |
| hello123 | :'( | 12:05 |
| hello123 | ikonia you know to setup raid 5? | 12:05 |
| hello123 | leitmedium? | 12:05 |
| ikonia | hello123: you need raid 5 ror raid 6 | 12:05 |
| ikonia | hello123: you need something that creates a raid array spanning your disks, rather than mirroring them | 12:05 |
| ikonia | I'd strongly advise against raid 0 for something that big though | 12:05 |
| hello123 | raid 5 | 12:05 |
| hello123 | raid 0 if 1 disk fails all gone | 12:05 |
| hello123 | thats also a problem | 12:05 |
| ikonia | exactly | 12:05 |
| ikonia | so you want raid 5 or 6 | 12:05 |
| hello123 | raid 5 | 12:07 |
| hello123 | ikonia can you setup raid 5? | 12:07 |
| xnox | hello123: raid installation instructions are here: https://help.ubuntu.com/12.04/serverguide/advanced-installation.html#software-raid | 12:07 |
| leitmedium | hello123: have you seen https://help.ubuntu.com/12.04/serverguide/advanced-installation.html = | 12:07 |
| leitmedium | hehe | 12:08 |
| xnox | snap! =) | 12:08 |
| hello123 | :p i just dont want to take risk as i am new to linux :( | 12:08 |
| hello123 | and even that how to delete | 12:08 |
| hello123 | the raid 1 array | 12:09 |
| hello123 | and make raid 5 | 12:09 |
| xnox | hello123: in the menus explained in that guide, you can add/delete raid arrays and choose any raid level (if you have enough drives) | 12:09 |
| hello123 | The mdadm utility can be used to view the status of an array, add disks to an array, remove disks, etc: | 12:10 |
| hello123 | not removal of array :( | 12:10 |
| hello123 | any1? | 12:13 |
| leitmedium | you have to "disassemble" it. | 12:14 |
| leitmedium | mdadm --stop /dev/md2 | 12:14 |
| hello123 | yeah thats good and ahead? | 12:15 |
| leitmedium | then follow the guide | 12:15 |
| hello123 | leit when i tried previous time it said sd1 device busy or unavaialble | 12:15 |
| hello123 | :( | 12:15 |
| hello123 | ? | 12:19 |
| leitmedium | hello123: is it still mounted? | 12:22 |
| hello123 | when i try to unmount /sda1 | 12:22 |
| hello123 | it says not mounted anywhere lol :p | 12:23 |
| bgardner | hello123: Did you actually try to umount '/sda1' or was it /dev/sda1? | 12:36 |
| hello123 | it was /dev/sda1 | 12:38 |
| bgardner | hello123: You should pastebin the output of mdadm --stop /dev/md2 | 12:40 |
| hello123 | @bgardner ok | 12:41 |
| bgardner | hello123: Let's keep the conversation here so others can weigh in and assist. | 12:41 |
| hello123 | yeah sure | 12:42 |
| merlijn| | hi, I was wondering if anybody is able to tell me what happened to the virtual kernel images in 14.04 - it appears you can only use generic kernels for this? | 12:46 |
| hello123 | how to change password of root from rescue mode? | 13:00 |
| ikonia | you shoudln't | 13:01 |
| bekks | root has no password to be changed. | 13:01 |
| ikonia | the root password shouldn't be set | 13:01 |
| hello123 | i cant login into server | 13:01 |
| hello123 | bt i can go into rescue | 13:01 |
| bekks | you cannot log in as root by default because root has no valid password set. | 13:02 |
| bekks | !root | hello123 | 13:02 |
| ubottu | hello123: Do not try to guess the root password, that is impossible. Instead, realise the truth... there is no root password. Then you will see that it is 'sudo' that grants you access and not the root password. Look at https://help.ubuntu.com/community/RootSudo | 13:02 |
| hello123 | psswd make a root passwrd right? | 13:02 |
| xnox | jamespage: will icehouse be available for 12.04 via cloud archive? | 13:02 |
| bekks | hello123: Please read the article linked by ubottu. | 13:03 |
| hello123 | ok | 13:03 |
| memoryleak | Hi | 13:12 |
| === marcoceppi_ is now known as marcoceppi | ||
| makara | when chrooting, why do we always mount /dev and /dev/pts separately? | 13:40 |
| sixBB | Hello. I have a question about mounting a thumbdrive. I type "sudo mount /dev/sde1 ~/thumbdrive" but when I do this, all the files are owned by root. Is there a way to avoid that? | 13:47 |
| DarkStompy | Is there any wany for me to reset a user password (as user, I am not an admin). I am unable to log in to the server. | 14:04 |
| sixBB | you can reset your own | 14:09 |
| sixBB | but if you're not an admin, you can't reset others' | 14:09 |
| sixBB | to reset your own, type "passwd" | 14:10 |
| DarkStompy | sixBB: Thanks. | 14:14 |
| sixBB | np | 14:14 |
| shredding | But you have to be logged in to change it ... | 14:14 |
| sixBB | yes. You have to know the old password even if you're already logged in. | 14:22 |
| sixBB | only an admin can change your password without knowing it | 14:22 |
| sixBB | but the user said that he was unable to login, not that he forgot his password. :D | 14:23 |
| === frankban_ is now known as frankban | ||
| === rcj` is now known as rcj | ||
| === Ursinha is now known as Ursinha-afk | ||
| === Ursinha-afk is now known as Ursinha | ||
| kirkland | jamespage: SpamapS_: rbasak: nice work ;-) http://mysqlrelease.com/2014/02/repos-and-distros-upstream-and-downstream/ | 15:28 |
| hello123 | hii | 15:31 |
| hello123 | bgardner there? | 15:31 |
| bgardner | hello123: Yes? | 15:32 |
| hello123 | yeah i got the server back what to check now? | 15:32 |
| bgardner | hello123: What is your question? | 15:33 |
| hello123 | make raid 5 all my 6 hdd | 15:34 |
| hello123 | 6 x 4tbhdd | 15:34 |
| bekks | Do you have a hardware raid controller? | 15:35 |
| hello123 | no softraid | 15:35 |
| === Ursinha is now known as Ursinha-afk | ||
| jamespage | kirkland, ta | 15:40 |
| hello123 | hello? | 15:44 |
| === Ursinha-afk is now known as Ursinha | ||
| === roo_ is now known as roo | ||
| webfox_ | HI, could someone help me to test and make it work my server network please? | 16:56 |
| webfox_ | I am trying to set a static ip address to it but still not being sucessful. | 16:57 |
| rbasak | kirkland: thanks! I love reading posts like that :) | 17:28 |
| kirkland | rbasak: ;-) | 17:31 |
| === SpamapS_ is now known as SpamapS | ||
| SpamapS | wait what | 17:49 |
| SpamapS | 5.6 in trusty? | 17:49 |
| SpamapS | that is.. | 17:49 |
| SpamapS | aggressive | 17:49 |
| SpamapS | kirkland: All I did was pawn all the work off on jamespage .. ;) | 17:50 |
| SpamapS | ok I see, 5.6 in universe still | 17:50 |
| SpamapS | not a good idea IMO | 17:50 |
| SpamapS | but hey, nobody asks me. :) | 17:50 |
| henkjan | 5.6 in trusty. cool! | 19:25 |
| hazmat | hallyn, i've noticed some oddity around lxc-clone destroying my container (btrfs snapshot) if its interuppted (13.10) leaving just a rootfs.hold and a missing rootfs directory... | 20:06 |
| hazmat | ie. its destroying the clone source if interuppted | 20:06 |
| hazmat | just curious if that sounds familiar. | 20:07 |
| hallyn | hazmat: no, please file a bug. we probably need to change the new container's rootfs in config earlier | 20:08 |
| hazmat | hallyn, sure.. github preferred for the bug? | 20:09 |
| hallyn | hazmat: lp would be good | 20:13 |
| thumper | stgraber, hallyn: at what process during boot do the lxc containers try to autostart? | 21:10 |
| thumper | is there any way I can make sure an upstart job happens before it tries? | 21:10 |
| hallyn | thumper: well you could do "start on starting lxc" | 21:11 |
| hallyn | autostart is done by /etc/init/lxc.conf which starts on runlevel 2345 | 21:12 |
| thumper | hallyn: so here is what we are planning... could be considered crazy | 21:14 |
| thumper | for fast lxc and the juju local provider we are wanting to have /var/lib/lxc be btrfs | 21:14 |
| thumper | we look, and if it isn't, we create a loopback device for it, and mount a btrfs in /var/lib/lxc | 21:15 |
| thumper | however if there are existing containers, we bail | 21:15 |
| thumper | and ask the user to either remove them or specify a directory override as a mount point | 21:15 |
| thumper | we'll have a process watch the free space and add extra space as necessary | 21:15 |
| thumper | FSVO magic | 21:15 |
| thumper | this way we can use the faster clone, and create juju template containers to clone from | 21:16 |
| thumper | also, we are making the juju autostart containers for the local provider to be an envron config option | 21:16 |
| thumper | but we need to support the case where the user has autostarting containers, but they are in the btrfs loopback magic fs | 21:17 |
| thumper | so I want to make sure we appropriately setup the loopback devices, and mount the btrfs in the right place before lxc tries to do the restart | 21:17 |
| hallyn | what about just mounting the btrfs elsewhere? I keep mine under /opt/lxc... | 21:18 |
| hallyn | they woulnd't autostart without a change, but you're making a change anyway | 21:18 |
| hallyn | (i suggest it bc then the user's existing containers under /var/lib/lxc wouldnt' matter) | 21:18 |
| thumper | hmm... we were considering that option | 21:19 |
| hallyn | i assume you've measured hwo btrfs on loopback performs? what about apt-get/dpkg on btrfs? | 21:20 |
| thumper | I'll be testing it | 21:20 |
| thumper | also | 21:20 |
| thumper | if we have a different directory for the containers | 21:20 |
| thumper | then we have to make sure that the user always specifies this dir when using the CLI tools | 21:21 |
| thumper | were hoping to avoid that... | 21:21 |
| thumper | but perhaps that is too much | 21:21 |
| thumper | hallyn: it seems that we can't pass a different root dir to lxc-autostart | 21:22 |
| thumper | in order to list the autostart containers | 21:22 |
| thumper | if we put them elsewhere | 21:22 |
| hallyn | thumper: lxc-autostart -P works | 21:23 |
| hallyn | should work | 21:23 |
| === unreal_ is now known as unreal | ||
| hallyn | thumper: anyway what you're suggesting sounds reasonable to me | 21:23 |
| thumper | oh, missing from the man page | 21:23 |
| hallyn | thumper: I assume you've considered lvm? | 21:23 |
| hallyn | yeah. stgraber: ^ autostart manpage needs to include the common opts page :) | 21:23 |
| thumper | hallyn: I'm taking what hazmat has and putting it into core | 21:24 |
| thumper | hallyn: he has good reasons AFAICT for btrfs | 21:24 |
| thumper | I really don't want to make it too complicated | 21:24 |
| thumper | hallyn: is lvm likely to be more efficient that btrfs on a loopback device? | 21:24 |
| thumper | I don't understand lvm | 21:24 |
| hallyn | thumper: yeah, it's only that on older kernels apt-get is *really* slow on btrfs | 21:24 |
| thumper | not looked at it at all | 21:25 |
| hallyn | thumper: no ths is not about the loopback device | 21:25 |
| thumper | hallyn: we are doing this for trusty + only | 21:25 |
| smoser | utlemming, | 21:25 |
| smoser | hey. | 21:25 |
| hallyn | oh. ok. there it's much better | 21:25 |
| smoser | cloud images have 'perl-modules' | 21:25 |
| thumper | o/ smoser | 21:25 |
| thumper | hallyn: yeah, I had concerns with older versions too | 21:25 |
| thumper | so I'm saying this is just a trusty feature | 21:25 |
| thumper | for the local provider | 21:25 |
| hallyn | thumper: so to be clear, you're goin to use an upstart job that starts on startling lxc andn does the logic you mentioned above right? | 21:25 |
| thumper | if I have a job that starts on start of lxc, will that happen before lxc runs? | 21:26 |
| xnox | thumper: btrfs on a loopback device, gains you none of the btrfs benefits. Neither does lvm on loopback device, much. Either reinstall, make your host be on top of lvm/btrfs. Or don't bother with lvm/btrfs. | 21:26 |
| smoser | utlemming, we have 'perl-modules', but perl-modules recommends each of | 21:26 |
| smoser | libarchive-extract-perl libmodule-pluggable-perl libpod-latex-perl libterm-ui-perl libtext-soundex-perl | 21:26 |
| smoser | but those are *not* in the image. | 21:26 |
| smoser | :-( | 21:26 |
| hallyn | thumper: yes. lxc wont' start until anything which starts on starting lxc completes | 21:27 |
| utlemming | smoser: hrm, interesting...I thought the recommends are installed. | 21:27 |
| utlemming | smoser: which release? | 21:27 |
| thumper | xnox: I have been told otherwise, and the only way to confirm is to test and measure | 21:27 |
| * thumper shrugs | 21:27 | |
| thumper | hallyn: how does clone work if not on lvm or btrfs? | 21:28 |
| thumper | hallyn: does it still work but slower? | 21:28 |
| hallyn | thumper: it either copies the whole directory, or uses overlayfs or aufs for snapshot | 21:28 |
| smoser | trusty. | 21:28 |
| hallyn | xnox: untrue though :) | 21:28 |
| xnox | thumper: you loopback file is stored on another filesystem, which is not checksummed nor snapshotable, while it appears that you have snapshots and checksuming, in fact you don't get that. | 21:28 |
| hallyn | xnox: a btrfs subvolume snapshot will still be much faster | 21:28 |
| smoser | utlemming, heres my list of "i don't understand" | 21:28 |
| smoser | http://paste.ubuntu.com/7007244/ | 21:28 |
| xnox | thumper: hallyn: if you want to snapshot loopback devices, convert them to be qemu-image based files which support efficient snapshots. | 21:28 |
| hallyn | no, subvolume snapshot is done at metadata level. it' | 21:29 |
| thumper | hallyn: ok, so worst case, lxc-clone on ext4 will just copy the root fs? | 21:29 |
| xnox | thumper: any host filesystem corruption, will propagate into all volumes snapshots in the loopback device. | 21:29 |
| hallyn | thumper: correct | 21:29 |
| thumper | hallyn: ok, ta | 21:29 |
| thumper | xnox: I appreciate that, and this is not what we are trying to use btrfs for | 21:30 |
| hallyn | thumper: so fallback to that is probably ok. it's a tradeoff | 21:30 |
| thumper | this is a dev tool | 21:30 |
| hallyn | if most of what you'll be doing is creating the container clones, then btrfs-on-lo will be good | 21:30 |
| thumper | hallyn: ok, I can test the speed behaviour of using clone independently of dealing with btrfs loopback | 21:30 |
| hallyn | if you'll be doing a lot of work on top of there, then not | 21:30 |
| thumper | so I can at least get some timing from that | 21:30 |
| hallyn | ok - ttyl | 21:30 |
| thumper | kk | 21:30 |
| hallyn | xnox: the main goal would be to avoid copying 300-700M of data on each container clone | 21:31 |
| smoser | hallyn, oh comon. IO is fun. | 21:31 |
| utlemming | smoser: I'll look into that shortly, I think this might be a live build issue | 21:31 |
| * hallyn hugs eatmydata | 21:32 | |
| smoser | utlemming, the only other thing in that list is python | 21:32 |
| smoser | i dont know where python3.3 is coming from in th eimages. | 21:32 |
| xnox | hallyn: dedup with hardlinks?! *har* *har* | 21:32 |
| hallyn | xnox: that sounds like the old vserver method | 21:33 |
| hallyn | which worked quite well | 21:33 |
| hazmat | thumper, actually aufs will be better | 21:33 |
| smoser | and yuck, looks like we have python3.3 and python3.4 in there! | 21:33 |
| thumper | hazmat: geez man, make up your mind | 21:33 |
| hazmat | thumper, its lxc 1.0 on trusty.. i'm giving it a whirl now.. | 21:33 |
| hazmat | thumper, well originally when we talked overlayfs several months ago.. i said i'd prefer aufs.. it just wasn't supported in lxc for this usage. | 21:34 |
| hallyn | hazmat: yeah i think stgraber can point you to any issues you'll run into with aufs. | 21:34 |
| hallyn | of course, there's the issue of potential regresions as aufs is afaik unsupported | 21:34 |
| hazmat | thumper, aufs vs overlayfs thread from last august fwiw | 21:35 |
| thumper | hazmat: ok, I'll defer working on any of the btrfs loopback bits | 21:35 |
| thumper | hazmat: and just move forwards with the template images and clone | 21:35 |
| thumper | hazmat: and we can measure differences from that | 21:35 |
| hazmat | thumper, cool. by big concern with btrfs.. is well dealing with size management is a pain on the loops as usage grows... | 21:35 |
| thumper | kk | 21:36 |
| hazmat | s/by/my | 21:36 |
| xnox | thumper: i'd be very surprised if you hit any bottlenecks. | 21:36 |
| thumper | getting templates, clone and a method to update templates will take me a while I guess | 21:36 |
| thumper | we can look at speeding up clone once those are in place | 21:37 |
| thumper | measure often | 21:37 |
| thumper | hazmat: sound reasonable? | 21:37 |
| hazmat | thumper, sounds good.. also we should talk auto apt-proxies at some point.. we had some for local in pyjuju days.. saved much bandwidth.. i didn't see it in core. | 21:37 |
| thumper | hazmat: it's there :-) | 21:38 |
| thumper | all that proxy work I did | 21:38 |
| hazmat | thumper, that's if you explicilty set one on the env.. but it doesn't set one up for you.. | 21:38 |
| hazmat | ? | 21:38 |
| thumper | FSVO of auot | 21:38 |
| thumper | local does | 21:38 |
| thumper | it looks to see what you have | 21:39 |
| thumper | and uses that | 21:39 |
| thumper | although it probably needs a tweak or two | 21:39 |
| hazmat | what you have on the host?.. but that means effectively its not being used.. | 21:39 |
| hazmat | ie. pyjuju installed apt-cacher-ng on host and configured containers for it.. | 21:40 |
| smoser | hazmat, you can't really do loopback devices. | 21:40 |
| hazmat | smoser, cause? | 21:40 |
| smoser | or at least if you're suggesting that you need to be aware that fsync doesn't work through them. | 21:40 |
| thumper | ok, perhaps I don't understand what you are after, or you aren't aware of what's happening, either way, can we shelve this for another time? | 21:40 |
| smoser | sometimes people care about fsync. | 21:40 |
| hazmat | smoser, hmm.. noted.. aufs is the preferred choice.. and thanks to stgraber is viable with lxc | 21:40 |
| hazmat | thumper, sure | 21:41 |
| thumper | hazmat: if we use aufs, how is that going to work if we update the base container from which it was cloned? | 21:41 |
| thumper | is there something we need to care about here? | 21:42 |
| thumper | o/ njpatel | 21:42 |
| hazmat | hmm.. its not going to like that.. | 21:42 |
| hazmat | thumper, tbd | 21:42 |
| thumper | hazmat: but we were wanting to keep the base template updated no? | 21:42 |
| hazmat | thumper, we may have to create new layers for updates | 21:42 |
| njpatel | 'ello there thumper | 21:42 |
| thumper | njpatel: what's up in SFO? | 21:43 |
| njpatel | stuff and things :) | 21:43 |
| njpatel | How's the cloud world? | 21:43 |
| === Ursinha is now known as Ursinha-afk | ||
| === Ursinha-afk is now known as Ursinha | ||
| === unreal_ is now known as unreal | ||
| mgw | Is there an option for apt to prevent it from trying to install a half-installed package "A" when I tell it to install package "B"? | 23:02 |
| sarnold | mgw: I think I've had to tell apt to uninstall A before trying to install B when I've faced that in the past | 23:04 |
| mgw | sarnold: thanks, not what I was hoping for | 23:05 |
| mgw | I'm working on the SaltStack aptpkg module, and it sort of blows up in that situation. | 23:06 |
| mgw | Was hoping there was an option that could just ignore those | 23:06 |
| sarnold | mgw: hrm. :/ maybe there's a dpkg --force- option that would work. I've tried to avoid those.. | 23:07 |
| mgw | sorry… I'm actually trying to remove package B | 23:07 |
| mgw | in this case | 23:07 |
| mgw | but I think an install would fail too | 23:07 |
| mgw | fail in the sense that apt-get would return non-zero | 23:07 |
| mgw | when it failed to configure package A | 23:07 |
| sarnold | mgw: oh, I wonder, can you just call dpkg --purge on B directly? | 23:09 |
| mgw | sarnold: no, it complains about dependencies | 23:14 |
| mgw | btw, in my test case it's Java related, so go figure | 23:15 |
| mfisch | zul: do you know if ceilometer-alarm-notifier is packaged for havana? | 23:17 |
| * mfisch will be back later to see the answer, commuting | 23:17 | |
| sarnold | mgw: hrm, is this related? https://bugs.launchpad.net/ubuntu/+source/openjdk-7/+bug/1273601 | 23:18 |
| mgw | sarnold: i don't think so | 23:19 |
| mgw | the underlying issue is with oracle-java7-installer | 23:19 |
| mgw | which fails to fully install | 23:19 |
| mgw | then when you try to install/remove other packages, it keeps trying to fix oracle | 23:20 |
| mgw | It's not that big a deal for me, I was just hoping there was an option I could add to the salt module to ignore that | 23:21 |
| sarnold | mgw: bugger. it was a stab in the dark anyway.. | 23:21 |
| mgw | sarnold: thanks | 23:22 |
| TJ- | mgw: There's a kludge I've used in the past, hacky but saved my sanity: edit /var/lib/dpkg/status, locate the "Package: " stanza, and change "Status:" to "install ok installed" temporarily | 23:25 |
| sarnold | TJ-: ha! nice. | 23:26 |
| TJ- | mgw: That example extracts the single package; to make it generate the entire new status file, you'd remove the final "M==1" conditional so it is just "{print $0}" | 23:41 |
| === SpamapS_ is now known as SpamapS | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!