Time Nick Message 12:59 * pdurbin reads - page allocation failure - am I running out of memory? - Server Fault - http://serverfault.com/questions/236170/page-allocation-failure-am-i-running-out-of-memory 13:08 pdurbin heh. on cleaning /tmp. :) http://cdn.memegenerator.net/instances/600x/28702594.jpg 14:03 pdurbin any tips on weaning myself away from gparted? would like to re-write my doc to use just parted instead when resizing a VM disk image 14:32 larsks pdurbin: Other than "don't use gparted"? What sort of tips are you looking for? 14:38 gridiron The only thing I remember from parted is apropos partition to show me what I needed 14:41 larsks Also, where's your doc? 14:46 pdurbin larsks: this doc needs work... rewriting it 14:47 pdurbin so, right now i boot off a parted magic cd and use virt-manager to do my thing. i wonder if i can boot off the parted magic iso from within `virsh console`... 14:55 larsks I was asking about the doc because I was curious what your current process looks like. When I resize an image, I typically extend the underlying image file on the host, run fdisk on the host to repartition, then reboot the guest instance and pvresize+lvresize. 14:57 pdurbin larsks: yeah, me too, but this guest has raw partitions. no LVM 15:08 larsks So instead of pvresize+lvresize, you just use resize2fs. The partition changes can still happen on the host using fdisk. 15:08 larsks ...or parted, of course, but same idea. 15:19 semiosis pdurbin: no node, just had to hack a few features into a sencha touch application over the weekend 15:20 semiosis was pleasantly surprised how easy it was 16:02 pdurbin semiosis: cool. hubot is supposed to be pretty hackable, being written in coffeescript and all 16:02 pdurbin larsks: right, could just use resize2fs... probably 16:20 pdurbin larsks: so / is mounted on /dev/sda5. i already ran `qemu-img resize /path/to/my-vm.qcow2 +10GB`. from the vm i should simply be able to run `resize2fs /dev/sda5` ? because i'm getting "The filesystem is already 2496256 blocks long. Nothing to do!" 16:20 pdurbin from here, i'm thinking i'll boot off a parted magic iso and run gparted (or parted) 16:26 larsks pdurbin: Generally, (a) resize disk image, (b) resize partition, and then finally (c) resize2fs. 16:27 larsks I run fdisk on the host after running qemu-img reisze. 16:27 larsks Without that step, the partition size hasn't changed --> the block device size hasn't changed --> resize2fs doesn't see any additional space. 16:28 pdurbin larsks: hmm, ok. what fdisk command, please? 16:30 larsks Are you looking for interactive or batch mode? Because so far I've only had to do it interactively, and I'm only resizing the last partition on the disks, so the sequence is: (d)elete the partition, add a (n)ew partition, and accept all the defaults. 16:31 pdurbin larsks: wait, in fdisk delete sda5 where all the data is?! 16:32 larsks Yes. Because you're creating a new partition starting at the same place and extending to the end of the disk. 16:32 pdurbin larsks: sounds a little scary 16:32 larsks It's the same thing as "modifying a partition", except you're more aware of what's happening. 16:33 pdurbin they call it parted magic for a reason :) 16:33 larsks It's an identical operation. Don't believe the magic, just understand the process. 16:33 pdurbin heh. i'm trying. thanks for helping :) 16:34 larsks Remember that deleting a parition in fdisk doesn't actual change anything on disk until you write out the changes...by which time you've created the new partition. 16:34 larsks So it really is just an update operation. 16:34 pdurbin larsks: yeah, that makes sense 16:51 pdurbin uh oh. "error: unknown filesystem. grub rescue" maybe i'll try parted for gparted... 16:51 pdurbin s/for/or/ 16:51 pdurbin i'm glad this is a vm :) 16:54 pdurbin virsh console doesn't work for grub :( 17:06 pdurbin huh. maybe i *do* need to convert this disk from qcow2 to raw... at least for the resize... gparted still sees the old size... 17:54 pdurbin qemu-img convert -f raw /tmp/my-vm-disk0.raw -O qcow2 /tmp/my-vm-disk0.qcow2 17:59 pdurbin (that's how i convert from raw to qcow2) 18:05 pdurbin larsks: have you ever seen this? http://michael.orlitzky.com/articles/resizing_a_kvm_or_qemu_disk_image.php 18:10 pdurbin ahh, good old gparted. so i definitely don't need to convert to raw. can keep it as qcow2 18:10 pdurbin maybe i'll take a peek at parted 18:11 pdurbin i wish i could run parted magic from within `virsh console` rather than virt-manager 18:55 pdurbin weird how pmagic-6.7.iso works for me but pmagic_2012_10_10.iso doesn't 19:22 larsks pdurbin: Had not seen that before. Given that you can just run qemu-img resize, it seems unnecessarily complicated. 19:34 pdurbin larsks: yes, `qemu-img resize` is what i use. plus gparted: increasing the size of a VM with fixed partitions (no LVM) - http://danceb.in/VIvPln0c4hGUBYyl9DNYg 19:34 larsks Yo, whatever works. 19:35 pdurbin i'd love to get away from gparted though. would prefer to use `virsh console` rather than virt-manager 19:35 larsks fdisk on the host + resize2fs! 19:35 larsks Hey, anyone here heading to DevOpsDays NYC? 19:35 pdurbin larsks: yeah.. that broke everthing. grub wouldn't boot 19:38 larsks Then you're probably doing some other than I described. 19:38 larsks But like I said, whatever works is the best solution. 19:38 pdurbin probably. i tried. oh well 19:40 pdurbin the devil is in the details :) 20:24 semiosis interesting -- http://internettrafficreport.com/event/2.htm 20:24 semiosis "The target of the attack were the 13 DNS root servers, which are responsible forhelping to resolving domain names to their respective IP's. Even though 9 of the 13 servers were disabled in the attack, the remaining were able to support the additional load without any widespead problems." 20:25 pdurbin semiosis: "Dateline: 10/22/2002"? 20:25 semiosis four of the root servers are so capable they can withstand an attack that knocks out *the other 9* and still keep up with the global demand for name resolution 20:25 semiosis oh 20:25 semiosis damn it 20:25 semiosis didnt have time to read 20:25 semiosis in any case, it was still epic, even though not today 20:26 shuff we even had DNS back in 2002? 20:26 semiosis ha 20:26 shuff i thought back then we were still rsyncing /etc/hosts 20:26 pdurbin those were the days 20:26 shuff when you knew your upstream gateway and your DNS servers IP addresses by heart 20:27 semiosis i mean, yeah, lets all celebrate the 10th anniversary of resiliant architecture standing up to evil attackers! 20:27 shuff so say we all 20:34 larsks pdurbin: https://gist.github.com/3933980 20:36 larsks That requires nothing but console access and will work for either qcow2 or raw images. 20:36 larsks It does have some caveats (which are noted in the document). 21:33 larsks Hey, remember what I said about being leery of salt rolling their own crypto? Now I own your salt cluster! https://github.com/saltstack/salt/issues/2239 21:35 semiosis wow and the reporter suggests running a decryption service to solve it 21:37 larsks I just run zeromq over stunnel when I want encryption and certificate verification.