| cpaelzer_ | good morning | 05:54 |
|---|---|---|
| === cpaelzer_ is now known as cpaelzer | ||
| === admcleod_away is now known as admcleod | ||
| === elsheepo is now known as beatzz | ||
| ahasenack | good morning | 11:21 |
| kstenerud | ahasenack o/ | 11:30 |
| ahasenack | hi kstenerud | 11:30 |
| frickler | can someone remind me of the URL for the version check matrix for UCA packages? somehow this has dropped from my list of bookmarks | 11:57 |
| ahasenack | cpaelzer: did you see my last comment on https://code.launchpad.net/~ahasenack/ubuntu/+source/cyrus-sasl2/+git/cyrus-sasl2/+merge/357779 ? Probably lost somewhere in your inbox storm | 11:59 |
| cpaelzer | ahasenack: I didn't see that yet | 12:26 |
| cpaelzer | hmm | 12:26 |
| cpaelzer | let me check if I see it in the ppa conf where I thought it was in the past | 12:26 |
| ahasenack | it's what prompted my emailing about bileto usage | 12:26 |
| cpaelzer | I'm sure I have seen valid xenial tests | 12:26 |
| ahasenack | well, they pass with other packages | 12:26 |
| ahasenack | but if you look there, they are also including the phone ppa | 12:27 |
| cpaelzer | oh, I see | 12:27 |
| ahasenack | it just so happens that none of the needed packages come from there | 12:27 |
| ahasenack | these two tests that are failing use qt packages | 12:27 |
| ahasenack | and apparently the phone has/had such packages at a higher version than xenial | 12:27 |
| cpaelzer | since most of the overlay was UI centric it makes sense that it is qt | 12:27 |
| cpaelzer | but OTOH most of our cases don't hit those bits | 12:28 |
| ahasenack | correct | 12:28 |
| cpaelzer | so unless most tests have issues I'd go on and consider this one an unlucky special case that needs manual tests | 12:28 |
| cpaelzer | as said on the mail thread | 12:28 |
| cpaelzer | but I'm still pretty much in favor of testing pre-upload | 12:28 |
| === Pici` is now known as Pici | ||
| smoser | https://code.launchpad.net/~smoser/usd-importer/+git/usd-importer/+merge/357826 | 15:08 |
| smoser | rbasak: ^ | 15:08 |
| rbasak | smoser: it is merged, but Launchpad is behind | 15:32 |
| rbasak | https://git.launchpad.net/usd-importer/log/ | 15:32 |
| rbasak | smoser: it should be in edge | 15:32 |
| smoser | hm. i'd never seen launchpad behind. | 15:42 |
| kklimonda | is there some documentation on running ubuntu core on bare metal servers, or is it not at that point yet, and still mostly focused on IoT? | 16:59 |
| ahasenack | cpaelzer: are you still looking at https://code.launchpad.net/~ahasenack/ubuntu/+source/cyrus-sasl2/+git/cyrus-sasl2/+merge/357779 ? | 17:01 |
| Guma | I just reinstalled my server and desktops to 18.04. I was wondering if there is a way to point all my desktops to my server for updates? The ideas is that I want all my desktops update to latest what is on server in order to keep my desktops on same level. | 17:53 |
| josefig | Hi, I have installed Ubuntu server 18.04 for LTSP server, I already installed everything but there's left the dnsmasq someone has information how to set it up with 2 NICs ? | 17:56 |
| lotuspsychje | Guma josefig idle here a bit ok, we having a bit of timezone glitch atm :p | 18:10 |
| josefig | hehe no worries, in the meanwhile i'm reading you know | 18:15 |
| josefig | I think it has to be with the second NIC, the first is connected to internet properly | 18:15 |
| === nacc_ is now known as nacc | ||
| RoyK | Guma: setup an apt mirror on the server and configure the desktops to use that server as an apt source | 18:25 |
| RoyK | Guma: or if your internet connection is good, just use unattended-upgrades to have them upgraded automatically as often as you wish | 18:26 |
| Guma | RoyK I never done this. Do you have some link to document waling me through it? Thank you so much | 18:26 |
| RoyK | Guma: first hit on google https://www.howtoforge.com/local_debian_ubuntu_mirror | 18:27 |
| Guma | Great. I will try. Thank you for getting me started | 18:28 |
| RoyK | good luck :) | 18:29 |
| Guma | One more thing. I have my own deb packages that I am creating. This is for company I am working on. They are private deb packages and also looking at hosting them privately | 18:29 |
| RoyK | Guma: again - google says https://wiki.debian.org/DebianRepository/Setup | 18:30 |
| Guma | Ok that will get me going. :) | 18:30 |
| RoyK | Guma: http://jfgi.herokuapp.com/images/bart.gif ;) | 18:31 |
| Guma | RoyK Fair enough. I always do. But in this case I was not sure where to start. Now I know. Cheers | 18:32 |
| sarnold | Guma: you can also set up a squid-deb-proxy if you don't want to spend a terabyte of disk space on a mirror | 18:39 |
| cpaelzer | ahasenack: I was fine with the cyrus fix given that you tested locally where the bileto test failed | 18:41 |
| ahasenack | cpaelzer: ok | 18:44 |
| JanC | nowadays a terabyte isn't all that much, of course :) | 18:47 |
| lotuspsychje | JanC: did you buy that 15tb WD? :p | 18:49 |
| JanC | no | 18:50 |
| josefig | If I'm trying to boot PXE and I get "DHCP packet received on enp3s0f1 which has no address" could be because of the network card has not IP assigned, even if I did a static declaration when I do a ip addr i don't see the ip. Why is that? or How is the correct procedure to do this ? | 19:24 |
| RoyK | josefig: do you have a dhcp server on that thing? | 19:28 |
| ahasenack | yeah, check if the server actually offered an ip | 19:28 |
| ahasenack | i.e., server logs are also useful | 19:28 |
| tomvolek | HI : I am trying to install Ubuntu from USB on a HP proliant server , I get a message "parition /de/sda1 assigned to / starts an offset 565248 bytes from minimum alignment for this disk" , is this a question for this group or the Ubuntu channel ? | 19:33 |
| tomvolek | Entire install disk is partitioned to one partition / and a swap . | 19:34 |
| sarnold | heh, that's a funny number.. 8192 * 69 | 19:36 |
| xnox | tomvolek, with UEFI or eithout UEFI? | 19:37 |
| tomvolek | sarnold: Some folks on the net mention you need to format the USB to make sure it does not use 512 byte sectors by default and increse it to 2048 , So I did that, used disk utility on mac to increase sectors on usb to 2048 so it would match linux. | 19:37 |
| xnox | tomvolek, also there should be either grub-bios slice inserted or an UEFI ESP partition. | 19:37 |
| tomvolek | xnox Sorry where does UEFI comes into play ? | 19:38 |
| tomvolek | xnox what do you mean by grub-bios slice inserted ? inserted where ? | 19:38 |
| xnox | tomvolek, well, if the server is booting using UEFI, you also need an ESP partition formatted as EFI. | 19:39 |
| tomvolek | I tried to keep thigns simple and created only one parition with swa | 19:39 |
| xnox | tomvolek, are you using automatic partitioning, or manual? | 19:39 |
| tomvolek | manual sir | 19:39 |
| xnox | tomvolek, and if you are booting using BIOS one needs to create a grub-bios partition (not sure if that is done automatically, even in manual mode) | 19:39 |
| tomvolek | on the ubuntu install screen, it allows you to reparition and I remove all previous partitions and recreate a root and a swap | 19:40 |
| xnox | well, without ESP (UEFI) or grub-bios (BIOS) you will not be able to install grub, and will fail to boot. | 19:40 |
| xnox | tomvolek, can you first try to use automatic partitioning to wipe and install? before playing with manual? | 19:41 |
| tomvolek | xnox, i certainly did not create a seperate parition grup-bios ... how much space should allocate tot that... I never had to do that on my other installs :) | 19:41 |
| tomvolek | xnonx .. sure, how do i do that ? | 19:41 |
| xnox | tomvolek, i know the details because that's what the backend code does. But i'm not sure if we actually show this in the UI or not. Cause it is fairly minor details, the grub-bios is very tiny, like a few kB after mbr | 19:42 |
| tomvolek | xnox.. I am booting from a usb stick , ubuntu 16.04 ... on an older HP DL 365 proliant .. | 19:42 |
| xnox | tomvolek, reboot installer, hit enter, until done. | 19:42 |
| xnox | tomvolek, at partitioning choose to "automatically partition the whole disk" or some such. | 19:42 |
| xnox | tomvolek, when done boot and checkout what partitioning layout is created, start/alignment, etc. | 19:43 |
| xnox | tomvolek, most likely first partitionin does not start at 0. And it's safer if one starts it at like 4MB to be honest. | 19:43 |
| xnox | (if one is manually partitioning things) | 19:43 |
| tomvolek | xnox, installer will detect whatever partition its on the disk already and present it, interesting it does not present me with option to partition automatically or manully, it displays the error I mentioned above and forces me to go back and try to delete the partions it detects on disk and recreate ....hence i get into this loop :) | 19:44 |
| xnox | ah | 19:44 |
| xnox | even when you reboot the installer, right? (it can mess up its own state machine) | 19:44 |
| tomvolek | its very intereting, i have installed lots of Ubunut , never seen this before | 19:45 |
| xnox | tomvolek, can you drop to shell, and delete things manually? the drive might have odd things leftover which installer fails to clean maybe? | 19:45 |
| xnox | like incomplete signatures of incomplete LVM groups, or parts of RAID, etc? | 19:45 |
| tomvolek | I do a hard reboot, recycle power on BIOS I set to boot from USB, it detects USB, boots inot instller . i take option to install, it asks for my time zone and then comes into this parition issue and stops | 19:45 |
| xnox | ctrl-alt-f4 should give you a terminal, or something | 19:45 |
| xnox | hm, sad. | 19:46 |
| tomvolek | xnox , ok i will try, I can go into a command line and start gparted ,, and remove all partitions, then reboot and come back into installer and see if this time i can do automatic partitioning ... | 19:47 |
| tomvolek | too much of my time is getting waisted on this one lab server :) | 19:47 |
| xnox | tomvolek, yeap. | 19:47 |
| ahasenack | sgdisk -Z <device>? | 19:47 |
| ahasenack | that erases it | 19:47 |
| tomvolek | ok let me try ..tx for suggestion .. stand by :) | 19:47 |
| xnox | if there is sgdisk available..... | 19:47 |
| xnox | tomvolek, also why use 16.04? and not 18.04? | 19:47 |
| xnox | the installer in 18.04 is well lit. | 19:47 |
| tomvolek | not sure, i am using ubuntu 16.04 stock on the USB | 19:47 |
| ahasenack | that is two years old | 19:48 |
| xnox | 18.04 has less screens, asks less things, does the installer quicker, and can even import ssh keys from launchpad or github. | 19:48 |
| tomvolek | xnox, reason is there is a bug in 18.04 with a graphich device driver :) , so i have to install 16.04 and then upgrade to 18.04 ... thats another long story :) | 19:48 |
| ahasenack | is this a desktop? | 19:49 |
| tomvolek | yes sir | 19:49 |
| xnox | oh | 19:49 |
| xnox | tomvolek, given the channel, i assumed you are trying to install Ubuntu Server | 19:50 |
| xnox | tomvolek, do you have internet? whilst you are in this broken state, could you please do $ ubuntu-bug ubiquity ? | 19:50 |
| xnox | and open a bug report, that would give me all the right logs. | 19:50 |
| xnox | tomvolek, and yeah sgdisk should work to unbreak you. | 19:50 |
| xnox | tomvolek, i typically use $ wipefs --all /dev/disk/by-id/* | 19:52 |
| tomvolek | xnox, i wasnt sure where to post my question thats why i asked at the start. Let me try sgdisk and report back . thanks guys | 19:52 |
| xnox | or use the disks app to format the drive. | 19:52 |
| xnox | tomvolek, it's all good =) | 19:52 |
| tomvolek | ok i back out of install and on the screen which says try Ubuntu , from here I should be able to get gparted or cmdtool started | 19:53 |
| tomvolek | FYI: this host has Raid 1+0 , two 146 G drives .. | 19:53 |
| xnox | hmmmm | 19:54 |
| xnox | tomvolek, but what raid is it? hardware raid? intel matrix raid? does it need drivers? | 19:54 |
| xnox | tomvolek, it's a bit odd installing desktop on a server =) | 19:54 |
| xnox | tomvolek, i would go the other way around, of using server installer, and then installing desktop with $ apt install ubuntu-desktop | 19:55 |
| tomvolek | xnonx ok thats a great suggestion, | 19:55 |
| xnox | let's just say desktop installer, might not be expecting raid devices =) | 19:55 |
| tomvolek | ok it has gdisk ... | 19:55 |
| tomvolek | could be ... | 19:55 |
| xnox | if you can bring networking up; you can use $ apt update; apt install | 19:56 |
| xnox | to get whatever you need. | 19:56 |
| tomvolek | yes, I am conected to network .. dhcp .. | 19:56 |
| tomvolek | what shoudl i say to gdisk options | 19:56 |
| tomvolek | gdisk -Z /dev/sda | 19:57 |
| tomvolek | interesting when I start gparted , it displays a message saying "The driver descriptor says teh physical block size is 2048 but Linux says its 512 bytes " | 20:00 |
| xnox | and /dev/sda ... is that usb stick you are booted from? | 20:02 |
| xnox | or the drive you are trying to install onto? | 20:02 |
| xnox | i typically check $ ls -latr /dev/disk/by-id/* | 20:02 |
| xnox | to make sure i know for sure who is who | 20:02 |
| tomvolek | @xnox @ahasenack : reporting back, I used gdisk , wiped out the hard drive, tried to reinstall, this time I selected LVM vs manually partitioning. Install is going forward at the moment, seemed like this worked :) thanks again for your suggestions. | 20:59 |
| Glorfindel | hey all, ubuntu server 16.04 running on my vps, I'm having issues with some of my applications not running, complaining of no storage space left. I have 50gb and ncdu shows 16gb has been used. is there a limit to what percentage of storage can be used in the /home/ dir? | 22:47 |
| Glorfindel | if I try to install anything it also complains of no storage space left... weird :/ | 22:50 |
| cryptodan_mobile | Run df -h | 22:51 |
| tomreyn | and df -i | 22:51 |
| cryptodan_mobile | Copy to dpaste.com share link | 22:52 |
| RoyK | Glorfindel: check cat /proc/sys/fs/file-max and cat /proc/sys/fs/file-nr | 22:53 |
| Glorfindel | cryptodan_mobile, tomreyn: here's each http://dpaste.com/0208Y7W | 22:54 |
| RoyK | some applications open a *lot* of files and thus the max can be reached | 22:54 |
| Glorfindel | RoyK: pretty new to linux, check them for what? | 22:54 |
| tomreyn | inodes are full on / | 22:54 |
| JanC | you might be out of inodes... | 22:55 |
| tomreyn | sudo find / -xdev -printf '%h\n' | sort | uniq -c | sort -k 1 -n | 22:55 |
| RoyK | better use xfs next time ;) | 22:55 |
| Glorfindel | so basically I have too many folders? | 22:56 |
| RoyK | folders are files | 22:56 |
| Glorfindel | *dirs | 22:56 |
| JanC | you can configure inodes when creating the FS | 22:56 |
| RoyK | an inode points to a files or a folder and you have three millions of them | 22:56 |
| JanC | the number of inodes | 22:57 |
| RoyK | usually 3M files won't happen with such a small fs | 22:57 |
| JanC | I don't think you can change it afterwards for ext3/ext4? | 22:57 |
| RoyK | you can't | 22:57 |
| tomreyn | you'd use the line i posted above to identify what consumes them mostly, then move those files to a new file system (optionally one with extra inodes) | 22:58 |
| JanC | it might be email or something like that... | 22:59 |
| Glorfindel | tomreyn: ahh, I see. I actually know what is using them up... the ls -R stream it generates almost lagged me out | 22:59 |
| Glorfindel | it's the tileserver I installed | 22:59 |
| JanC | ah, that seems a possibility too if it stores each tile in a separate file | 22:59 |
| tomreyn | it should probably not be on /, and maybe be backed by a DB. | 23:00 |
| JanC | a filesystem is a database too :) | 23:00 |
| tomreyn | if a file system then most likely a separate file system with different mount options and extra reserved space | 23:01 |
| Glorfindel | ok, so if I understand this correctly; an inode is what stores the data about what files to show if a user is looking in a specific directory, right? and I ran out of those somehow? | 23:01 |
| JanC | no | 23:01 |
| RoyK | Glorfindel: what are you storing on this to get that amount of inodes? | 23:01 |
| Glorfindel | RoyK: I am hosting a minetest server, and I installed a tileserver that reads the minetest database and generates a real-time map of the world | 23:02 |
| Glorfindel | and it created a lot of directories to store tiles in I guess, not 100% on that | 23:02 |
| RoyK | Glorfindel: ok - so millions of small files? | 23:02 |
| Glorfindel | yeah | 23:02 |
| JanC | inodes describe the file | 23:02 |
| RoyK | (or dir) | 23:03 |
| tomreyn | or other file system object | 23:03 |
| JanC | multiple names in multiple directories can point to the same inode | 23:03 |
| Glorfindel | so it's metadata? | 23:03 |
| JanC | e.g. in case of hardlinks | 23:03 |
| JanC | part of the metadata for a file, yes (not the filename) | 23:03 |
| RoyK | Glorfindel: the "database" that holds filenames etc | 23:03 |
| Glorfindel | be back in a few minutes | 23:04 |
| JanC | inodes don't contain filenames | 23:04 |
| JanC | directories contain filenames :) | 23:04 |
| RoyK | JanC: and directories are inodes, since they're basically files containing the names of the files inside them | 23:05 |
| RoyK | well, inodes point to directories | 23:05 |
| JanC | right, the latter | 23:05 |
| RoyK | yes | 23:08 |
| JanC | well, in theory on-disk "inodes" could contain filenames on some filesystems if they refer to a directory that contains only a small number of files where they store the file (directory) data in the inode record itself, I guess | 23:12 |
| RoyK | doesn't matter much for Glorfindel - (s)he needs a new filesystem to make this work - I'd guess XFS would be better - no hard limits on the inodes | 23:15 |
| sarnold | ouch running out of inodes on / | 23:15 |
| JanC | either that or set the inode limit for ext* high enough | 23:15 |
| sarnold | "this is a usenet server thanks" | 23:16 |
| JanC | and probably best to store such data on a separate filesystem indeed | 23:16 |
| RoyK | JanC: and preferably not on ext4 :þ | 23:16 |
| sarnold | doubleyes :) hehe | 23:17 |
| JanC | well, you can set the limits on ext4 if you want | 23:17 |
| JanC | but using XFS or such might be easier | 23:18 |
| sarnold | these files have an average size of ~5k.. given how huge the binaries are going to be, I assume the actual data files are likely to be <4k.. | 23:18 |
| RoyK | dd if=/dev/zero of=somexfsfs.img bs=1M count=10k ; mkfs -t xfs … | 23:19 |
| sarnold | lol | 23:19 |
| sarnold | (seriously that's probably a good way out of this) | 23:20 |
| JanC | not going to work if you have no inodes left ;-) | 23:20 |
| RoyK | that's why I wrote that ;) | 23:20 |
| RoyK | JanC: well, you can probably release one inode for the image file | 23:21 |
| RoyK | and another for the mount point | 23:21 |
| sarnold | probably there's a useless kernel image.. :) | 23:23 |
| JanC | and hopefully mkfs doesn't need temporary files ;) | 23:23 |
| JanC | or something like that | 23:23 |
| RoyK | JanC: I doubt it - otherwise - just delete some old log files | 23:24 |
| RoyK | it should be doable | 23:24 |
| Glorfindel | I can release 3 million of them... | 23:25 |
| Glorfindel | and regenerate the tileserver stuff after | 23:25 |
| RoyK | then do so | 23:25 |
| JanC | if you can, (re-)install the VPS with a separate filesystem for that stuff | 23:26 |
| RoyK | Glorfindel: but you'll need a filesystem supporting a lot of more inodes for that than what you hve now | 23:26 |
| RoyK | JanC: if the provider allows you to choose fs… | 23:26 |
| Glorfindel | JanC: at this point I'm going to make that the last solution, as I have quite a bit of data and no good place to migrate it to | 23:27 |
| Glorfindel | well this is the longest I've seen rm take to delete anything | 23:27 |
| RoyK | it'll take some time to remove 3M files | 23:27 |
| sarnold | yeah rm isn't going to go quick :) | 23:28 |
| JanC | well, if you can create a new VPS for a month or so and copy everything over... | 23:28 |
| RoyK | but you can just background it and start creating an image file and putting xfs on it | 23:28 |
| Glorfindel | RoyK: can you explain what that command you sent earlier will do? | 23:28 |
| sarnold | (if it were ZFS, then you could delete the dataset, but that would also have required a different setup from the start, too. :) | 23:28 |
| RoyK | Glorfindel: the dd command just creates an empty file, 10GB in size - the mkfs -t xfs, creates a filesystem | 23:29 |
| JanC | well, booting from ZFS would be even more of a problem maybe | 23:29 |
| RoyK | Glorfindel: after it's created, mount it somewhere, add it to /etc/fstab and it should be fine | 23:29 |
| Glorfindel | ah, so it would be like a virtual disk? | 23:29 |
| RoyK | Glorfindel: yes | 23:30 |
| RoyK | JanC: I've used ZFS for almost a decade - not at home anymore, though - I prefer the flexibility of md ;) | 23:30 |
| Glorfindel | JanC: also I don't think this vps provider gave me the option for what fs to use | 23:30 |
| JanC | booting a linux VPS from ZFS, I mean | 23:31 |
| JanC | Glorfindel: you can't install a custom OS on it? | 23:31 |
| RoyK | JanC: I doubt most VPS providers support ZFS on the guest | 23:31 |
| JanC | well, it's Ubuntu, so ZFS support is included | 23:31 |
| sarnold | yeah pretend I didn't mention zfs.. it was just an aside that *some* filesystems offer way faster ways to remove three million files than rm :) | 23:32 |
| Glorfindel | JanC: it offers ubuntu 16.04 or windows server (forget what year) for an additional fee | 23:32 |
| JanC | the problem is more the booting, depending on the type of VPS this is | 23:32 |
| Glorfindel | may have been other distros too, I don't remember | 23:32 |
| Glorfindel | I'll check | 23:32 |
| RoyK | Glorfindel: just create that big, empty file and place a filesystem on it | 23:33 |
| JanC | my VPS provider allows me to install my own OS if I prefer (they won't give support on how to do that though) | 23:33 |
| Glorfindel | I was misremembering, you can install other disrtos via iso | 23:34 |
| RoyK | never mind zfs - it's not what you need here | 23:34 |
| Glorfindel | RoyK: "dd if=/dev/zero of=somexfsfs.img bs=1M count=10k ; mkfs -t xfs"? | 23:35 |
| RoyK | # dd if=/dev/zero of=somexfsfs.img bs=1M count=10k | 23:35 |
| RoyK | yes - just put that img file under /var/something | 23:36 |
| RoyK | somewhere it may belong | 23:36 |
| RoyK | perhaps /var/lofs | 23:36 |
| RoyK | or whatever | 23:36 |
| Glorfindel | any reason to not put it in ~/ ? I only ask because it's easier for me to remember the path if I keep everything in /home | 23:37 |
| RoyK | then mkfs -t xfs /path/to/that/file | 23:37 |
| Glorfindel | s/put it in/mount it in/ | 23:37 |
| RoyK | place it where you like, but filesystems don't really belong in /home | 23:37 |
| JanC | using an image file might have a small impact on FS speed though... | 23:37 |
| Glorfindel | ok | 23:37 |
| RoyK | JanC: it will, but it'll work | 23:37 |
| RoyK | Glorfindel: if you can get a new vdisk for the VPS, it'll probably be best | 23:38 |
| JanC | well, depending on how important FS speed is (maybe not if this is just for some personal test setup) | 23:38 |
| Glorfindel | RoyK: I just added an additional 20gb of ssd storage as "block storage" from 1and1 | 23:39 |
| Glorfindel | I'll check and see where it got put | 23:39 |
| JanC | ah, 1and1 | 23:39 |
| RoyK | Glorfindel: lsblk? | 23:39 |
| Glorfindel | rm is still going at it lol | 23:39 |
| Glorfindel | RoyK: sdb | 23:39 |
| Glorfindel | that was easy, thanks | 23:40 |
| RoyK | well, put xfs on it | 23:40 |
| JanC | Glorfindel: that also depends on how directories are structured :) | 23:40 |
| Glorfindel | so mkfs -t /dev/sdb ? | 23:41 |
| RoyK | mkfs -t xfs /dev/sdb # or perhaps put lvm on it first - just in case | 23:41 |
| Glorfindel | how do I put lvm on it and what is the purpose? | 23:42 |
| RoyK | !lvm | 23:42 |
| ubottu | Tips and tricks for RAID and LVM can be found on https://help.ubuntu.com/community/Installation/SoftwareRAID and http://www.tldp.org/HOWTO/LVM-HOWTO - For software RAID, see https://help.ubuntu.com/community/FakeRaidHowto | 23:42 |
| Glorfindel | thanks | 23:43 |
| sarnold | bah are those links actually usfeul here? | 23:43 |
| JanC | I wonder if/how filesystems optimize stuff like removing all files from a directory... | 23:43 |
| RoyK | Glorfindel: lvm is a nice abstraction layer on top of a 'physical' device - just vgcreate myvolumegroup /dev/sdX ; lvcreate -l 100%FREE -n myvolume myvolumegroup | 23:43 |
| RoyK | JanC: xfs does it quite quickly - ext4 doesn't | 23:44 |
| JanC | you'd need something like temporarily disabling index updates which some databases support | 23:44 |
| RoyK | zfs has a massive backlog so it seems to do it quickly, but yet, it takes some time | 23:44 |
| JanC | right, it would say ready, and then start doing the work ;) | 23:45 |
| JanC | (well it would record the fact that it will do that first) | 23:45 |
| JanC | still doesn't seem easy though, as the application uses a syscall for each file, I assume? | 23:46 |
| JanC | unless there is a syscall to delete a whole tree? | 23:47 |
| RoyK | JanC: the way ext4 works, you'll need a systemcall or five for each file, yes | 23:47 |
| RoyK | I doubt such a systemcall exists | 23:48 |
| RoyK | perhaps in zfs - certainly not in ext4 | 23:48 |
| JanC | it's useless if rm doesn't use it anyway | 23:49 |
| JanC | it seems like something that would be useful in some cases... | 23:49 |
| Glorfindel | oh hey it finished | 23:49 |
| Glorfindel | now maybe I can tab-complete | 23:50 |
| RoyK | :) | 23:50 |
| Glorfindel | ok, so should I put lvm on it? do the benefits outweigh the complexity? | 23:50 |
| RoyK | Glorfindel: did you stop the process holding those files open? | 23:50 |
| Glorfindel | no, but it crashed when it ran out of inodes | 23:51 |
| Glorfindel | yesterday | 23:51 |
| RoyK | I'd say put lvm on it | 23:51 |
| RoyK | but that's up to debate, I guess | 23:51 |
| JanC | can you resize that block storage without losing data? | 23:52 |
| Glorfindel | JanC: I just added it half an hour ago, there isn't any data on it yet | 23:52 |
| JanC | it might already be using something like lvm on the supervisor level | 23:52 |
| JanC | Glorfindel: I mean, if you ever want to in the future | 23:52 |
| RoyK | the only issue with xfs AFAIK is that it can't be shrunk | 23:53 |
| RoyK | but then - who shrinks a filesystem? | 23:53 |
| Glorfindel | ohhh.. no, I don't think so. when I look at the level I'm at in regards to resources they don't allow downsizing plans due to storage size changing | 23:53 |
| JanC | well, that might be because they can't guarantee your filesystem can downsize ;) | 23:54 |
| RoyK | Glorfindel: anyway - try 'vgcreate somevolumenme /dev/sdb' | 23:54 |
| Glorfindel | seems lvm is installed, but the command didn't work as I didn't sudo | 23:55 |
| JanC | but if they down allow upsizing either, lvm would allow you to increase the filesystem over 2 (or more) such "block storage" devices | 23:55 |
| JanC | *if they don't allow* | 23:56 |
| RoyK | JanC: or PVs | 23:56 |
| JanC | the "block storage" they provide to the VPS would be the PVs, of course | 23:57 |
| JanC | not sure they provide _real_ _physical_ volumes | 23:57 |
| RoyK | probably some hierarchical storage | 23:58 |
| RoyK | keep the old data on spinning rust and sell it all as 'ALL SSD' | 23:58 |
| sarnold | :) | 23:58 |
| JanC | might be storage on some SAN too | 23:58 |
| RoyK | you don't get hierarchical storage on anything but on a SAN | 23:59 |
| RoyK | there have been some OSS projects, but I don't think any has landed yet | 23:59 |
| JanC | oh yes you do; you can buy SSHDs everywhere ;) | 23:59 |
| RoyK | not quite the same ;) | 23:59 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
