=== nhandler_ is now known as nhandler === nhandler_ is now known as nhandler === alkisg1 is now known as alkisg [10:01] hello, could anyone help with a problem with a thin client not booting due to nbd_server problems? [10:01] thx [12:27] stgraber: please do the following changes to the ltsp packaging: [12:27] * debian/ltsp-client-core.install:client/initramfs/hooks/ltsp usr/share/initramfs-tools/hooks [12:27] - debian/ltsp-client-core.install:client/initramfs/scripts/ltsp_nbd usr/share/initramfs-tools/scripts [12:27] - debian/rules:cp $(CURDIR)/client/initscripts/ltsp-setup $(CURDIR)/debian/ltsp-client-core.ltsp-client-setup.init [12:27] ...we won't be using ltsp_nbd and ltsp-setup anymore :) [12:27] (and if vagrantc wants he can completely remove ltsp-setup) [12:28] Ah and: - rules: dh_installinit --name=ltsp-client-setup --no-start -u"start 32 S ." [14:13] alkisg: ok, I'll have a look [14:50] alkisg: so should I remove all of these you listed above? [14:51] stgraber: yes [14:51] The line with the star * is a modification [14:51] hooks/ltsp_nbd renamed to hooks/ltsp [14:51] alkisg: right, I did that one already I think (in the latest PPA build) [14:52] So the known problems for now are: [14:52] 1) broken default SCREEN_07, I'm fixing it now, [14:53] 2) no nbd-proxy wrapper, will do that next, [14:53] 3) borked shutdown, we'll need again upstart jobs for reboot/shutdown, but not the old ones, new ones to do nbd-client -d; echo o > sysrq trigger etc [14:53] So that no nbd-server processes are left on the server [14:54] As now that nbd-server is not started from inetd, it takes 2 hours for the connections to get killed [14:54] ok, I can restore the old ones and add nbd-client -d [14:55] Right, although it would be nice if they could somehow wait a bit until the ssh + epoptes connections are closed, I'll see if I can find a clean way for that [14:56] Killing epoptes should do it, not sure about ssh [14:58] pkill -f '^epoptes-client \+m' [15:02] alkisg: http://paste.ubuntu.com/841781/ [15:02] oops [15:03] http://paste.ubuntu.com/841784/ [15:03] /dev/mapp will be replaced by /dev/nbd btw [15:04] http://paste.ubuntu.com/841786/ [15:04] stgraber: hmmm if nbd-proxy is killed, then nbd-client will cause a disk access error, no? [15:05] probably indeed, I guess I'll need to move it to a separate kill after nbd-client [15:05] Also, exec reboot won't work unless you cache it first, better use the REISUB magic [15:05] And I'm using `exec -a` to run epoptes-client, so killall won't work, please use the pkill I wrote above [15:06] echo o > /proc/sysrq-trigger [15:06] I've tested that that works even without a root nbd connection active [15:08] alkisg: http://paste.ubuntu.com/841792/ [15:09] I think killall isn't available in thin chroots, let me see if I have one... [15:09] alkisg: I guess these two should be moved upstream actually and applied to the overlay at boot time [15:09] stgraber: ah, yes, right [15:09] So leave them to me, I'll put them to init-ltsp.d/common [15:09] or /ubuntu, better [15:11] alkisg: because of an overlayfs bug, you'll need to call 'initctl reload-configuration || true' after writing the jobs to /etc/init/ [15:11] alkisg: or actually, after doing any change to /etc/init [15:11] stgraber: err why? It appears to work fine so far [15:11] We do that in the init=/sbin/init-ltsp phase, right? [15:12] ah right, doing it before starting init works indeed, sorry ;) [15:12] :) [15:12] the bug only appears when init is already started [15:12] Yeah I would worry if we were to do that [15:13] So once I commit the default screen fix (really soon), ltsp will be bootable again, so a new package will be nice. [15:13] ok, just ping me when you need it [15:13] ty [16:42] stgraber: I fixed many problems, I think it's time for a new package if you please. [16:42] Next goal, nbd-proxy-wrapper :) [16:43] alkisg: ok, uploading one now [16:43] ty [16:43] alkisg: next time you push something to trunk, can you do "bzr tag current --delete" before "bzr push"? [16:44] alkisg: current shouldn't exist in the LP branch, it's just a local tag I use for snapshots ;) but it requires a commit to drop it [16:45] Sure [16:45] In about an hour, along with the nbd-proxy wrapper script [16:48] alkisg: uploaded === nhandler_ is now known as nhandler === alkisg1 is now known as alkisg