=== doko_ is now known as doko [08:03] popey: Looks like those packages you were asking about were auto-synced last night. [08:08] super, thanks cjwatson [09:00] Could someone approve update-manager 1:0.156.14.15 in precise-proposed? [10:39] hmmm, how's the libav transition going? [10:41] cjwatson, ^ would you have time to review/approve update-manager in precise today? we need it for testing hwe eol notifications. [11:21] jibel: doing [11:48] cjwatson, thank you [13:02] cjwatson: Hi! You are the SRU vanguard today, correct? [13:36] cjwatson: If you saw my previous ping, no need to respond. [14:25] http://people.canonical.com/~ubuntu-archive/component-mismatches.html new toy [14:25] (also priority, architecture) [14:26] also the next build of http://people.canonical.com/~ubuntu-archive/nbs.html should have a graph [14:28] if anyone knows how to make YUI's numeric axes work more sensibly, let me know [22:27] infintiy, cjwatson: are either of you around? [22:28] infinity, cjwatson: due to bug #1336855 and the recent Grub SRU for 12.04 anyone running an HVM instance on AWS will lose their instance on reboot. [22:28] bug 1336855 in cloud-init (Ubuntu) "[SRU] non-interactive grub updates for 12.04 break on AWS" [Critical,Confirmed] https://launchpad.net/bugs/1336855 [22:45] utlemming: fun. [22:46] infinity: yup...and if there is a grub update for 14.04, they'll get burned too [22:46] infinity: do you have any guidance how to dig out of this hole? [22:47] infinity: I am leary to put the logic into cloud-init since cloud-init doesn't have a grub dependency [22:52] utlemming: Erm, but the bug is in cloud-init? [22:53] infinity: right. So what happens is one first boot cloud-init sets a debconf selection for grub-pc "install_devices". Since it doesn't know to look at /dev/xvda and /dev/xvda, grub tries to install to /dev/sda (which is what the images originally had). [22:53] utlemming: Isn't it as simple as making that 1-line change to cc_grub_dpkg.py? [22:54] infinity: nope. Because that module only runs once. [22:54] infinity: and then there is the race condition of a user who updates grub and then cloud-init [22:54] utlemming: Then the bug log seems to lie. [22:55] infinity: adding a better, clarifying comment [23:01] infinity: clarified the bug [23:03] utlemming: I'm confused how you got in this state in the first place. Wouldn't have the first run of cloud-init created a broken grub config straight up and this has been wrong since day 1? [23:04] infinity: the cloud image build process installs grub by hand, since we can't build the images in AWS. [23:04] utlemming: That doesn't really answer my question. :) [23:05] utlemming: Any kernel or grub upgrade would have broken this, so what introduced the bug, and how long has it been broken? [23:06] infinity: that part I am unsure of. But AWS just reported it. It has been broken since day one. Not having looked, I would guess that grub changed something such that the bits installed on disk don't work with the stuff in /boot/grub. [23:10] utlemming: So, if we really wanted to make this smooth, we could SRU cloud-init with the 1-line fix, and have it conditionally re-run 'cloud-init-cfg grub-dpkg' in its postinst on upgrade from known-broken versions, *and* re-upload a grub SRU with Breaks: cloud-init (<< new_version), forcing cloud-init to upgrade before grub. [23:11] utlemming: I'm not sure having this leak into the grub packaging is really worth the hassle, but if you always want upgrades to work, I'm not sure there's a cleaner way. [23:11] infinity: I _did_ start to go down that path, but had an issue with the fact that 'cloud-init-cfg grub-dpkg' runs debconf-set-selection, which runs into a debconf lock [23:12] utlemming: Oh, I guess you could re-run grub-install on cloud-init postinst in the same condition that you run cloud-init-cfg grub-dpkg [23:12] utlemming: There should only be a lock if you're running it in a part of your postinst that happens to be debconfy? [23:13] infinity: the cloud-init postinst is debconf'd, yes [23:13] utlemming: Sure, but not the WHOLE postinst. :) [23:13] utlemming: If nothing else, there's some convenient imaginary whitespace before or after the bits that are. [23:13] infinity: lol, okay, I'll look at doing that