smoser | slangasek, around ? | 00:46 |
---|---|---|
smoser | http://paste.ubuntu.com/1191848/ | 00:46 |
smoser | so it seems like 2 issues. a.) network is already up, so i think ifup decides it doesn't need to bring the network up, and b.) mountall doesn't like me doing mounts for it. | 00:46 |
slangasek | smoser: how do you mean, "network is already up"? | 00:55 |
smoser | iscsi root. | 00:55 |
smoser | it came up in the initramfs | 00:55 |
slangasek | smoser: right; ifupdown doesn't consider that "up" | 00:55 |
smoser | hm.. | 00:56 |
slangasek | to ifupdown, it's only up if ifup has been called | 00:56 |
smoser | i'd have thought. | 00:56 |
slangasek | which is not to say that the network already being up can't be causing problems for ifupdown | 00:56 |
slangasek | or maybe iscsi even injects state? | 00:56 |
smoser | injects state? | 00:57 |
smoser | hm.. | 00:58 |
slangasek | smoser: yes, as in, perhaps the open-iscsi initramfs script scribbles something in /run to tell ifupdown to under no circumstances bounce the interface | 00:58 |
smoser | i dont think so. | 00:58 |
smoser | but clearly it cannot bounce the interface. | 00:58 |
smoser | but i dont know what i thought was wrong. about mountallnot liking me | 01:00 |
slangasek | smoser: dunno... there's some noise about filesystems already mounted, but coincidentally these are all optional mounts | 01:02 |
slangasek | smoser: looks like mountall itself succeeds, it's just the network not happening | 01:03 |
smoser | right. | 01:03 |
smoser | it was the "busy" stuff that i was confused by. | 01:03 |
smoser | i'll look foranything iscsi might be doing to protect | 01:04 |
smoser | or maybe ifupdown somehow just "knows"that it has non-local filesystems | 01:04 |
smoser | although it appears to have tricked mountal | 01:05 |
slangasek | smoser: should I be seeing full upstart logging here? there's nothing after 37.584470, at which point udev has not actually succeeded in starting and therefore ifupdown is not being triggered | 01:05 |
slangasek | smoser: upstart-udev-bridge is 'start on starting udev'; it never completes | 01:05 |
slangasek | at least, not that's shown in this log | 01:07 |
smoser | well, it still hasnt then (looking at that log now) | 01:09 |
slangasek | smoser: what's 'status upstart-udev-bridge' show? | 01:10 |
smoser | hm.. start/running | 01:11 |
slangasek | ok | 01:11 |
slangasek | so maybe the log output went somewhere else | 01:11 |
slangasek | (in which case... we probably need to see that log output, wherever it is, to get the whole picture) | 01:11 |
smoser | yeah. | 01:12 |
smoser | i have had thoughts that i was not getting all of output | 01:12 |
smoser | /var/log/syslog: http://paste.ubuntu.com/1191875/ /var/log/kern.log: http://paste.ubuntu.com/1191876/ | 01:18 |
xnox | anything interesting in /var/log/upstart/* ? | 01:19 |
xnox | cause that's where it's going by default.... | 01:20 |
smoser | $ ls -l /var/log/upstart -d | 01:22 |
smoser | ls: cannot access /var/log/upstart: No such file or directory | 01:22 |
slangasek | smoser: um? where'd you put that directory then? It's shipped by the upstart package | 01:40 |
smoser | well... i changed the timeout waiting for static-networking-up back to 120 seconds (i had put it to 10) in cloud-init-nonet.conf | 01:41 |
smoser | http://paste.ubuntu.com/1191899/ | 01:41 |
smoser | well, i didn't put that directory anywaere. | 01:41 |
smoser | suck. | 01:41 |
smoser | so this is definitely a bug. some cleaning out of images. in the build process kills that. | 01:41 |
smoser | so i'll open a bug on that. | 01:41 |
smoser | i changed that timeout back to 120 seconds | 01:42 |
smoser | and http://paste.ubuntu.com/1191889/ is the result | 01:43 |
smoser | its more obvious that there something is still blocking, waiting for that. | 01:43 |
smoser | probably for mounted / to finish | 01:43 |
slangasek | smoser: so http://paste.ubuntu.com/1191875/ shows a correct udev-based bring-up of the networking | 01:45 |
slangasek | but the timing is a bit odd | 01:46 |
* slangasek looks at the latest | 01:46 | |
slangasek | smoser: hah; udevtrigger waits to find out if you're in a container | 01:47 |
slangasek | smoser: /etc/init/container-detect.conf has an explicit check for mounted MOUNTPOINT=/run instead of virtual-filesystems | 01:47 |
slangasek | smoser: try twiddling it to use virtual-filesystems instead | 01:47 |
smoser | bah. | 01:47 |
slangasek | smoser: that seems like a change we can easily make in the distro - if it works | 01:48 |
smoser | well, it might be by design. | 01:48 |
smoser | in that it can run earlier than virtual-filesystems (much earlier) | 01:48 |
slangasek | nah | 01:48 |
smoser | altougho i guess nothing (as we see) is gonna guarantee that anyway. | 01:49 |
slangasek | I mean, it is, but there's nothing in the design that requires the rest of mountall to wait for the container answer | 01:49 |
slangasek | and changing the container-detect job is better than making your workaround job double-emit the mounted event | 01:49 |
smoser | so.. https://bugs.launchpad.net/ubuntu/+bug/1047707 is to cover /var/log/upstart in the images. | 01:49 |
ubottu | Launchpad bug 1047707 in Ubuntu "/var/log/upstart deleted from cloud images" [High,Triaged] | 01:49 |
smoser | well i had my work around in place there. | 01:50 |
smoser | and i think its required to get to the point. | 01:50 |
infinity | smoser: I assume that's a symptom of a larger "purging stuff willy-nilly" bug? | 01:50 |
infinity | smoser: Unless you were explicitly removing that one directory for some reason... | 01:50 |
smoser | yeah, it is. | 01:50 |
slangasek | smoser: yes, and your workaround does NOT emit the 'mounted' event, it emits virtual-filesystems | 01:50 |
smoser | no. its like rm -Rf /var/log/* | 01:50 |
smoser | right. | 01:50 |
infinity | smoser: Gah. | 01:50 |
slangasek | which we should bring container-detect into alignment with | 01:50 |
smoser | right. | 01:50 |
smoser | one thing that i'm running into | 01:50 |
smoser | the cloud-init-nonet job has 'output console' | 01:51 |
smoser | but its output is *not* getting to console | 01:51 |
smoser | where console=/dev/ttyS0 (per that command line). | 01:51 |
smoser | i had to explicitly | tee /dev/ttyS0 to get it to the console. | 01:51 |
smoser | i'll try the change on container and boot | 01:52 |
smoser | slangasek, that seems to fix the proble. | 02:08 |
* slangasek nods | 02:09 | |
smoser | well, with the hack-emit-virtual-filesystems. | 02:09 |
slangasek | it certainly should have been the last thing left that could be a problem :) | 02:09 |
smoser | so i can get where i need to go now, i think. | 02:09 |
smoser | so you want a bug against upstart for container-detect? | 02:10 |
slangasek | smoser: can you file a bug on upstart about container-detect and /run vs. virtual-filesystems? | 02:10 |
slangasek | yes please :) | 02:10 |
smoser | i'll open one, but i'd appreciate if you fill in some better debug and justfication. | 02:10 |
smoser | so that some time from now when i'm trying to trace steps i'll have some info | 02:10 |
slangasek | smoser: ack | 02:12 |
smoser | https://bugs.launchpad.net/ubuntu/+source/upstart/+bug/1047712 | 02:14 |
ubottu | Launchpad bug 1047712 in upstart (Ubuntu) "container-detect.conf should be 'start on virtual-filesystems'" [Undecided,New] | 02:14 |
smoser | slangasek, what woudl you think about trying to make mountall happier by running my virtual-filesystems-emitter on 'starting mountall' | 02:16 |
smoser | (in addition to cloud-init-nonet. it should just prep the system so mountall doesn't htink it should have to mount those filesystems) | 02:16 |
smoser | slangasek, well. one more piece of my puzzle (i'm stillnot getting resolv.confworked out). | 02:53 |
smoser | but /etc/init/iscsi-network-interface.conf is a bit of info | 02:53 |
smoser | http://paste.ubuntu.com/1191973/ | 02:55 |
slangasek | smoser: 'start on starting mountall' should be fine for that, yeah | 03:13 |
slangasek | smoser: right, that's kinda what I was expecting to see from iscsi | 03:13 |
slangasek | and it's skipping the ifupdown hooks, which is a problem | 03:14 |
slangasek | though from the comment, it seems that may have been deliberate | 03:14 |
smoser | slangasek, yeah. i'm poking around. | 03:21 |
smoser | i'm in initramfs now. | 03:22 |
smoser | was hoping i could get it to store off the dns server responses | 03:22 |
smoser | from the dhcp request. | 03:22 |
smoser | they get printed. and "configure_networking" pretends that "Ip-Config tries to create this file and when it succeds" (/run/net-"${DEVICE}".conf) | 03:23 |
smoser | but it sure doesn't look like it does to me. i'm in an initramfs and dont see any such stuff getting created. | 03:23 |
smoser | shoot. | 03:26 |
smoser | it loo like our initramfs is busted in that respect | 03:26 |
smoser | it writes /tmp/net-eth0.conf | 03:26 |
smoser | at least in precise | 03:26 |
slangasek | so, where does ipconfig come from? busybox? | 03:27 |
smoser | klibc | 03:27 |
smoser | http://paste.ubuntu.com/1191999/ | 03:28 |
slangasek | smoser: fixed in quantal; seems klibc-utils and initramfs-tools were out of sync on this in precise | 03:30 |
slangasek | smoser: sounds like that should be cherry-picked from klibc-utils | 03:30 |
smoser | ah. so the klibc-utils updated to /run | 03:30 |
slangasek | klibc 2.0-2, in Debian | 03:30 |
slangasek | yes | 03:30 |
smoser | i'm opneing a bug for that. | 03:30 |
smoser | easiest thing might be to just update to lok in 2 places | 03:31 |
slangasek | no, just update klibc to use /run | 03:31 |
slangasek | that's the right place for it to be | 03:31 |
slangasek | and nothing relies on the current behavior in precise | 03:31 |
smoser | slangasek, are you sure this is "fix-released" ? | 03:35 |
smoser | https://bugs.launchpad.net/ubuntu/+source/klibc/+bug/1047722 | 03:35 |
ubottu | Launchpad bug 1047722 in klibc (Ubuntu) "configure_networking net-DEVICE.conf in /run, but ipconfig writes in /tmp" [Undecided,New] | 03:35 |
slangasek | smoser: I've checked the source in quantal, it's definitely writing to /run onw | 03:36 |
slangasek | now | 03:36 |
smoser | k. | 03:36 |
smoser | so i thin i shoudl be able to get that information from the dhcp response (dns info) | 03:38 |
smoser | and store it somewhere | 03:38 |
smoser | and then replay it to resolvconf | 03:38 |
slangasek | in fact, so long as you're getting it written to /run, I believe the initramfs should be mount -omove'ing that for you | 03:40 |
smoser | yeah, i wondered about that. so i dont have to store it even. | 03:42 |
slangasek | yep | 03:42 |
slangasek | you do have to make sure you're not mounting something else over /run ;) | 03:42 |
smoser | i think i'm gonna close this laptop now. | 03:42 |
smoser | you've been very helpful, slangasek | 03:42 |
smoser | thanks. | 03:42 |
slangasek | y/w | 03:42 |
slangasek | have a good evening | 03:42 |
smoser | evening for 8 more minutes here | 03:43 |
smoser | then am | 03:43 |
smoser | later. | 03:43 |
slangasek | looking at the dictionaries-common code makes me very angry | 03:56 |
=== cpg is now known as cpg|away | ||
=== cpg|away is now known as cpg | ||
=== cpg is now known as cpg|away | ||
=== cpg|away is now known as cpg | ||
=== wzssyqa1 is now known as wzssyqa | ||
=== wzssyqa is now known as Guest43660 | ||
malkauns | can someone point me to how i can create animated progressbars in gtk-3? | 07:31 |
=== iulian is now known as Guest54251 | ||
=== Guest54251 is now known as iulian | ||
=== mcclurmc_away is now known as mcclurmc | ||
=== cpg is now known as cpg|away | ||
=== korn_ is now known as c_korn | ||
Daviey | jtaylor: thanks for doing asterisk | 11:10 |
jtaylor | np but it needs doing for precise | 11:11 |
jtaylor | unfortunately it are ~8 CVEs and the simple diff are 10000 lines ... | 11:11 |
jtaylor | I have no time for that and no idea how to test anything | 11:11 |
Daviey | jtaylor: Well, if you are able to prepare an update.. I'm happy to test it? | 11:13 |
jtaylor | thanks, I'll come back to you if an update gets prepared | 11:16 |
Daviey | super! | 11:17 |
=== doko_ is now known as doko | ||
penguin42 | doko: Which gcc package version were you using for bug 1019725 ? | 13:51 |
ubottu | Launchpad bug 1019725 in gcc-4.6 (Ubuntu) "tiny STL code sample segfaults cc1plus" [Medium,Triaged] https://launchpad.net/bugs/1019725 | 13:51 |
=== yofel_ is now known as yofel | ||
pcarrier | is it just me, or is canonical trying to get rid of the gnome ecosystem and quantal a step toward more and more components in qt? | 15:07 |
ScottK | Any particular examples in mind? | 15:22 |
alkisg | I'm trying ubuntu-defaults-image, and it's working ok except that the generated .iso doesn't have the following files: | 15:34 |
alkisg | autorun.inf boot dists install pics pool preseed README.diskdefines ubuntu wubi.exe | 15:34 |
alkisg | ...can I do something to include those as well? | 15:34 |
phaidros | is there any reason that 12.04 LTS comes without CONFIG_CIFS_ACL kernel setting ?! | 16:14 |
=== cpg|away is now known as cpg | ||
penguin42 | jamespage: Thanks for fixing the iPXE | 19:15 |
jamespage | penguin42, np | 19:16 |
penguin42 | jamespage: What's your setup like; I'm running two kvm vm's one with tgt, and I've got two different clients; one disk less, and the other with disk - the diskless install doesn't seem to have setup the inittab properly yet (from last weeks iso) | 19:17 |
jamespage | penguin42, I run tgt on the KVM host alongside libvirt and I use diskless kvm vms with a scripted ipxe setup todo automated iscsi root installs for testing | 19:18 |
jamespage | penguin42, its a bit hacky but it works - https://code.launchpad.net/~james-page/+junk/iscsi-testing | 19:19 |
penguin42 | jamespage: Have you had any problem with boot order issues on tgt? I seem to need to restart tgt after server boot for it to pick the disk up | 19:22 |
jamespage | penguin42, I'd not noticed anything like that - but my test cases don't normally persist between boots :-) | 19:23 |
jamespage | I don't make the tgt configuration persistent | 19:23 |
jamespage | penguin42, BTW have you tried 12.04 against a quantal tgt target? I'm wondering whether I need to SRU that fix to 12.04 as well | 19:24 |
penguin42 | jamespage: No I haven't; I haven't got a 12.04 host to try | 19:24 |
penguin42 | (I guess I could go nuts and go nested kvm...) | 19:25 |
jamespage | penguin42, thats my prob as well | 19:25 |
jamespage | I probably will try that; I have access to a cloud which supports it | 19:25 |
penguin42 | jamespage: I've done nested kvm before - it's surprisingly easy; other than getting the default netmask right, it just works if you stack 12.04's | 19:25 |
penguin42 | jamespage: I suspect the other thing the people running iscsi in the real world would want to do automatically in the installer is get multipath going | 19:30 |
jamespage | penguin42, I have lots of experience with multipath in fibre channel based SAN's - but zip with iscsi | 19:33 |
penguin42 | jamespage: I don't know much about it - my server setup does have two net interfaces and I've bound it to both 'portals' - and in the non-root version I've got I have it log in once on each interface and see both instances as separate discs in /proc/scsi/scsi so I assume multipath would sort it out into one device | 19:35 |
penguin42 | jamespage: I think like FC there is a lot of 'magic' knowledge of timeouts and stuff with particular vendors | 19:36 |
penguin42 | jamespage: The only other commonality is that multipath is renowned as a pain | 19:37 |
=== cpg is now known as cpg|away | ||
=== ml|vacation is now known as ml|transit | ||
penguin42 | jamespage: Bug 1047998 is my problem with the server install on iscsi root | 20:42 |
ubottu | Launchpad bug 1047998 in debian-installer (Ubuntu) "iscsi root: iscsi not started in initramfs" [Undecided,New] https://launchpad.net/bugs/1047998 | 20:42 |
penguin42 | jamespage: and bug 1048008 for tgt needing a restart | 21:02 |
ubottu | Launchpad bug 1048008 in tgt (Ubuntu) "tgt needs restart to find LUN" [Undecided,New] https://launchpad.net/bugs/1048008 | 21:03 |
pcarrier | ScottK: ubuntu-sso-client-qt, ubuntuone-control-panel-qt, apparently libaccounts | 21:06 |
=== jtechidna is now known as JontheEchidna | ||
=== jtechidna is now known as JontheEchidna | ||
=== cpg|away is now known as cpg |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!