| smoser | harmw, yeah, i wanted to get one before i leave for holiday. | 10:51 |
|---|---|---|
| smoser | there is an 0.3.2~pre2 (which is not really released at http://download.cirros-cloud.net/local/0.3.2~pre2/) | 10:52 |
| smoser | i've not done any testing on it though, so i didn't even release it as a "pre2" yet. | 10:52 |
| smoser | so any feedback/test you could give it would be good. | 10:59 |
| harmw | ok, well i'll see what I can do tonight | 11:19 |
| harmw | there are some good fixes, iirc not that hard to verify | 11:20 |
| harmw | smoser: just me or is that ~ in the url causing bigtime 404 errors | 14:44 |
| harmw | ah, no | 14:46 |
| harmw | forgot the /local | 14:46 |
| smoser | yeah. | 14:46 |
| harmw | shouldn't it instant-reboot after resizing /? | 14:53 |
| harmw | i'm seeing hostroutes, excellent (so yes, metadata was fetched aswell since my hostname is correct) | 14:53 |
| harmw | hm, after reboot it hasn't grown to 3GB (the volume size) | 14:55 |
| harmw | or am I completely wrong and is growing root a step I specifically had to set/configure? | 14:55 |
| === Daviey_ is now known as Daviey | ||
| === ctracey|away is now known as ctracey | ||
| med_ | harmw, you may need to reinvoke smoser if you wanted some feedback on that 3G growth question | 16:49 |
| med_ | I suspect after two hours he looked away | 16:49 |
| smoser | harlowja_away, it should grow the roto filesystem . i think. | 16:50 |
| smoser | i can't imagine that that was broken. | 16:50 |
| smoser | i thikn you're talking about cirros, right? | 16:51 |
| smoser | it should have done it. | 16:51 |
| med_ | is harmw the same as harlowja_away ? | 17:00 |
| smoser | shoot. no. | 17:51 |
| smoser | harmw, ^ | 17:51 |
| === harlowja_away is now known as harlowja | ||
| harlowja | haha | 17:54 |
| smoser | hm.. harmw cirros *should* grow the root partition and resize2fs it also. you should probably see something on console output if it didnt. (console output as in 'nova console-log') | 18:17 |
| harmw | good one med_ , though i wasn't particularely after a quick-reply :) | 18:20 |
| med_ | heh. | 18:20 |
| harmw | and yes smoser, it should've done that though I didn't see it happen | 18:20 |
| harmw | let me try again | 18:21 |
| harmw | (on a different cloud) | 18:21 |
| smoser | harmw, you're booting the disk.img file ? | 18:22 |
| harmw | yea, converted to raw and then it goes to glance | 18:24 |
| harmw | i remember having seen a syslog line telling it resized, just cant remember when that was | 18:24 |
| smoser | you don thave to convert to raw, harmw | 18:27 |
| harmw | without that I can't copy-on-write, and glance doesn't (yet) do it for me | 18:27 |
| smoser | well glance shouldn't convert for you | 18:28 |
| smoser | glance image-create --disk-format=qcow2 --container-format=bare --name=smoser/cirros-0.3.2~pre2-amd64-disk.img --file ~smoser/src/cirros/mirror/0.3.2~pre2/cirros-0.3.2~pre2-x86_64-disk.img | 18:28 |
| smoser | nova should convert for you. | 18:28 |
| smoser | nova should see a qcow and convert it to raw and then create qcow backed by the raw. | 18:28 |
| harmw | isn't nova supposed to just use whatever cinder throws at it? in my case, a glance image that originates in rbd (ceph) | 18:29 |
| harmw | growpart: mode: auto devices: ['/'] | 18:29 |
| harmw | i have that in my userdata script, amongst some other stuff | 18:30 |
| harmw | though that would probably not be needed to have cloudinit autogrow, right? | 18:30 |
| smoser | oh. i see. | 18:32 |
| smoser | cinder. you did say that. boot from volume. | 18:32 |
| smoser | yeah. you're right. | 18:33 |
| harmw | yes, sorry | 18:33 |
| smoser | so get the cuos yeah, i would have expected that to work, but have not tested that. | 18:33 |
| smoser | so what steps do you do to acocmplish that ? | 18:33 |
| smoser | * download image, convert to raw. | 18:33 |
| smoser | how did you then upload / register a boot-from-volume image? | 18:34 |
| smoser | i dont know if i hcan do this with the cloud i have here. | 18:34 |
| smoser | harmw, http://paste.ubuntu.com/6584812/ | 18:34 |
| smoser | thats patebin of non-volume boot. and you see line 266 there. | 18:35 |
| harmw | yes, the GROWROOT:bla line is what i've seen before | 18:35 |
| smoser | there could be race conditions there. | 18:35 |
| smoser | :-( | 18:35 |
| harmw | ah | 18:35 |
| harmw | well, i didn't debug it | 18:35 |
| harmw | perhaps i can do that later on | 18:36 |
| harmw | let me paste the commands i'm using first | 18:36 |
| harmw | http://paste.ubuntu.com/6584832/ | 18:39 |
| smoser | oh. maybe i can do this. i'll try | 18:39 |
| smoser | harmw, bah. i can't really test. | 18:46 |
| smoser | the cinder create for me causes error | 18:46 |
| harmw | i'm back in a few minutes | 18:56 |
| smoser | harmw, i did this: http://paste.ubuntu.com/6585009/ | 19:07 |
| smoser | taking 'path 2' there. (attach a volume and populate it myself) | 19:07 |
| smoser | then console-log http://paste.ubuntu.com/6585013/ | 19:08 |
| smoser | so it *could* be racy still, but it seemed to work here. | 19:08 |
| smoser | harmw, hm.. so i see it has failed in my instance here. | 19:14 |
| harlowja | openstack, racy, not possible, lol | 19:27 |
| harmw | GROWROOT: CHANGED: partition=1 start=16065 old: size=64260 end=80325 new: size=6265350,end=6281415 | 19:31 |
| harmw | weird | 19:31 |
| med_ | harlowja, openstack is racy. Danica Patrick is our new sportscar driver. | 19:35 |
| harlowja | :) | 19:36 |
| smoser | harmw, yea, so it does do the partition resize. that works. | 19:36 |
| harmw | indeed, but that wasn't on my console this afternoon | 19:36 |
| smoser | but then the call to 'resize2fs' which happens from | 19:36 |
| smoser | /etc/init.d/resizefs | 19:36 |
| harmw | console also tells me the hostname is cirros | 19:36 |
| smoser | seems to return failure with no error to stderr. | 19:37 |
| harmw | hm, it didn't set the hostname | 19:38 |
| harmw | whre does cloud-init keep its log in cirros? | 19:42 |
| harmw | hm, looks like it received the public-hostname, though hostname is empty (in /var/run/cirros/datasource/data | 19:43 |
| smoser | harmw, thats odd. mine get their hostnames. | 19:46 |
| harmw | mine normally do to :) | 19:46 |
| smoser | what does 'hostname' say ? | 19:50 |
| harmw | hostname said the right thing | 19:50 |
| harmw | i've just booted a new instance, all is ok with this one :| | 19:50 |
| harlowja | smoser is metswin:) still the login for cirros? | 19:50 |
| harlowja | or was that cubwin | 19:50 |
| harmw | cubs | 19:51 |
| harlowja | nice, ha | 19:51 |
| harmw | this new instance has a proper hostname, GROWROOT did something aswell but resizefs didn't do squat and fdisk -l does not show a fully used disk either | 19:54 |
| smoser | harmw, fdisk -l doesn't ? | 19:58 |
| smoser | that is odd. | 19:58 |
| smoser | what does cat /proc/partitions say ? | 19:58 |
| smoser | harmw, so it works in 2 portions | 19:58 |
| smoser | growroot runs in iinitramfs | 19:58 |
| smoser | and changes the parttion table | 19:59 |
| smoser | then unmounts the target | 19:59 |
| smoser | err.. | 19:59 |
| smoser | unmounts | 19:59 |
| smoser | grows | 19:59 |
| smoser | remounts | 19:59 |
| smoser | then pvirot root | 19:59 |
| smoser | then /etc/rc3.d/S55-resizefs | 19:59 |
| smoser | *should* resize the filesystem "per-once" | 19:59 |
| smoser | and i see this functional on "ephemeral store" images. | 20:00 |
| harmw | /dev/vda1 * 16065 6281414 3132675 83 Linux | 20:10 |
| harmw | $ cat /proc/partitions | 20:10 |
| harmw | major minor #blocks name | 20:10 |
| harmw | 253 0 3145728 vda 253 1 3132675 vda1 | 20:10 |
| harmw | i've now manually run /etc/rc3.d/S55-resizefs start, and after thats done df doesn't show any change | 20:16 |
| harmw | though after another second or so it does show the changed fs, as beeing bigger | 20:17 |
| smoser | harmw, right. it backgrounds. | 20:20 |
| smoser | that is by design. (possibly bad design) | 20:21 |
| harmw | hmk | 20:21 |
| smoser | but it does so because resize2fs in cirros is pretty slow. | 20:21 |
| harmw | aha | 20:21 |
| smoser | at least compared to 12.10 ubuntu or later. | 20:21 |
| harmw | ok | 20:21 |
| smoser | and the thoguht was that people want to run scirors on very un-performant systems | 20:21 |
| smoser | such as non-kvm qemu | 20:22 |
| harmw | what bugs me is it didn't resize on initial boot, only just now when I did it manually | 20:22 |
| smoser | yeah. that was confusing to me too. | 20:22 |
| smoser | i even put debug into that /etc/rc3.d/S55-resizefs | 20:23 |
| smoser | to run 'cat /proc/partitions' | 20:23 |
| smoser | and output was correct. | 20:23 |
| smoser | hm.. wait a minute. i think i just saw it fail on ephemeral root | 20:24 |
| smoser | when it fails, you'll see a file /tmp/resize.out | 20:25 |
| smoser | that should have stderr and stdout of the 'resize2fs /dev/root' | 20:26 |
| smoser | but it doesnt | 20:26 |
| harmw | yea | 20:26 |
| harmw | $ cat /tmp/resize.out | 20:26 |
| harmw | resize2fs 1.42.2 (27-Mar-2012) | 20:26 |
| harmw | and thats it | 20:26 |
| === ctracey is now known as ctracey|away | ||
| harmw | well, new instance (again): growroot did its trick according to consolelog but resizefs only put the above in the logfile | 20:47 |
| smoser | harmw, its really odd. | 20:48 |
| smoser | cirros-per doesn't even put its marker down | 20:48 |
| smoser | which it should do even on failure. | 20:48 |
| === ctracey|away is now known as ctracey | ||
| harmw | smoser: its not putting down the marker file, indeed, it only shows this: | 22:18 |
| harmw | using state_dir=/var/lib/cirros/sem, marker=once.once.resizefs | 22:18 |
| harmw | [once] sh -c o=/tmp/resize.out; | 22:18 |
| smoser | yeah, its just wierd. | 22:18 |
| harmw | followed by resize2fs [..] | 22:18 |
| smoser | i dont get it. | 22:18 |
| harmw | me neither | 22:18 |
| smoser | it must be something stupid | 22:18 |
| harmw | yea, well even after a reboot it didn't resize - so growfs isn't causing the pain | 22:21 |
| harmw | ill see again, tomorrow | 22:21 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!