12:25:48 #startmeeting virt_test_day_f18 12:25:48 Meeting started Thu Nov 1 12:25:48 2012 UTC. The chair is crobinso. Information about MeetBot at http://wiki.debian.org/MeetBot. 12:25:48 Useful Commands: #action #agreed #halp #info #idea #link #topic. 12:34:16 what does this mean? 12:34:18 libvir: XML-RPC error : authentication failed: polkit\56retains_authorization_after_challenge=1 12:34:21 Authorization requires authentication but no agent is available. 12:34:24 libguestfs: error: could not connect to libvirt (URI = NULL): authentication failed: polkit\56retains_authorization_after_challenge=1 12:34:27 Authorization requires authentication but no agent is available. [code=45 domain=7] 12:43:59 rwmjones: there isn't any polkit agent listening in your session. so for gnome3 the agent is gnome-shell, otherwise its usually polkit-gnome 12:44:04 rwmjones: how are you triggering it? 12:44:24 I have: 12:44:25 /usr/lib/polkit-1/polkitd --no-debug 12:44:28 rwmjones: the agent is the bit that actually asks for your password 12:44:34 running, but no polkit-gnome 12:44:45 I've no idea how, it just started happening today 12:46:25 rwmjones: try running /usr/libexec/polkit-gnome-authentication-agent-1 maybe it's failing for some reason. but generally your desktop session autostarts that 12:46:26 hmm...I just upgraded to F18 and am also hitting polkit issues. 12:46:54 I'm being asked for the root password to manage libvirt domains, 12:47:23 the config file I have in /etc/polkit-1 doesn't seem to be working anymore. 12:47:52 there are some bits about polkit and a couple bugs here: https://fedoraproject.org/wiki/Upgrading_Fedora_using_yum?rd=YumUpgradeFaq#Fedora_17_-.3E_Fedora_18 12:48:41 * stefanha tries 12:48:50 rwmjones: are you using gnome shell, xfce, ...? 12:48:58 stefanha: ^ same question 12:49:05 gnome-shell 12:49:15 xfce 12:49:34 this all worked fine until very recently 12:49:51 Same here 12:50:27 libvirt - polkit is still borked for me 12:50:39 (after polkit/systemctl reinstall and restart as described on the wiki) 12:51:13 "error : remoteDispatchAuthPolkit:2838 : Policy kit denied action org.libvirt.unix.manage from pid 3565, uid 1000: exit status 3" 12:51:17 I filed a bug for my problem here: https://bugzilla.redhat.com/show_bug.cgi?id=872166 12:52:46 rwmjones: that output is coming straight from polkit btw 12:53:04 stefanha: can you pastebin your /etc/polkit-1 file 12:54:05 crobinso: http://paste.stg.fedoraproject.org/1403 12:54:57 Perhaps I can run polkit in debug mode. It doesn't seem to log anything so I don't know why my policy file isn't working anymore. 12:55:33 stefanha: yeah, killall polkitd, then run it by hand with no flags and it prints debug output to stderr 12:56:53 stefanha: ah, I know now. polkit totally changed their rules file format. it's now javascript :( 12:57:33 stefanha: I've got an example somewhere, one sec 12:58:21 stefanha: http://pkgs.fedoraproject.org/cgit/openstack-nova.git/tree/nova-polkit.rules 13:01:21 crobinso: I replaced the file with the openstack javascript snippet (s/nova/stefanha/) but it still rejects me. 13:01:34 Going to paste the logs 13:02:25 crobinso: http://fpaste.org/We8w/ 13:02:45 Still no mention of my policy file. Perhaps I need to rename it or put it somewhere else. 13:03:51 hmm...yeah. I wonder if /etc/polkit-1/localauthority still exists in F18 :) 13:04:32 stefanha: I think that's it :) I stuck my copy in /etc/polkit-1/rules.d and it worked 13:05:08 stefanha: btw does that fpaste link actually work for you? it gives me 500 internal errors 13:05:17 crobinso: bingo. Thanks for your help 13:05:40 crobinso: Works every now and then. I get 500 sometimes too. 13:05:45 ;-) quality 13:06:28 I'm surprised that polkit changed their config file format without no backwards compatibility 13:06:58 There's a high chance that anyone who has custom polkit rules is now broken. 13:08:31 stefanha: yep! pretty lame IMO. particularly with libvirt where it's common for people to have custom rules 13:10:16 rwmjones: can you repost your libvirt polkit policy patch from a few weeks back? I think it'll get ack'd now 13:12:54 What is the F18 way to regenerate an initramfs and rerun GRUB2 config generation? 13:13:08 I seem to have a new F18 kernel without initramfs 13:13:17 So I had to boot into the old F17 kernel temporarily. 13:14:21 stefanha: not sure about that, someone in fedora-kernel should know 13:15:00 crobinso: my patch? 13:15:12 oh you mean the one changing the defaults 13:15:15 let me find that 13:15:23 rwmjones: that's the one 13:20:22 sent 13:30:48 does anyone know if qemu:///session is a valid URL if you're running as root? 13:30:59 should that start a sort of private root-only libvirtd? 13:33:27 rwmjones: I think that should work 13:49:44 Hello - I'm here to help out with the test day. I'm in the process of getting my own machine set up to run tests under F18, but I can also answer questions anyone else may have, especially related to snapshots 13:53:10 hi eblake :) 14:04:01 eblake: do you know if libvirt supports qemu xbzrle migration compression? 14:08:49 crobinso: not yet 14:09:05 I know there's an open BZ for adding support though 14:09:36 https://bugzilla.redhat.com/show_bug.cgi?id=842857 14:13:39 I'm here to answer any usb-redir questions. Mainly the usb 1/2 spice usbredir stuff. Not the usb-3 host redir stuff, although I'm interested in that too 14:20:29 when is the virt-test-day starting? 14:23:06 zer0c00l: It's on 14:33:46 eblake, just a side question: would there be any more fixes going into libvirt-git before tomorrow ? 14:34:12 until 1.0.0 is out, we are at bug fix only 14:34:21 I do know of a couple bugs that I'd like to fix, and today may turn up more 14:34:35 for example, the new virNodeGetCPUMap API needs to work on RHEL 5 14:34:59 but Peter's external system checkpoint series will not go in until after the release 14:35:46 ok, thanks, 1.0.0 is slated tomorrow, right? 14:36:09 and alas, that means it won't be in F18; and even worse, it probably won't even be in rawhide or fedora-virt-preview until 1.0.1 in another month, which means it won't get as much testing as I would have liked 14:36:28 yes, 1.0.0 is slated for tomorrow - the 7-year anniversary of the first libvirt release 14:36:30 I'm not looking for external system checkpoint, but, just wondering if I should grab rpms from today's git, or rc2 (though it shouldn't matter for my demo) 14:37:05 the difference between rc3 and today's git is so far pretty small; and probably won't matter for your demo 14:37:20 it might be nice to use tomorrow's 1.0.0 release, if you can afford to wait 14:37:26 yeah, thx for that. 14:37:47 eblake: I'm having trouble remembering - did you say that you're upgraded an F17 box to F18 with yum? or from DVD? 14:37:50 um, I'll be on a plane tomorrow . I can try after getting down though 14:37:54 eblake: well I'm hoping to get virt-manager snapshot UI in rawhide before the end of the year, and then it'll probably get a ton of testing :) 14:38:28 I'm in the middle of updating a 32-bit F17 into F18 using yum 14:38:37 DVD won't upgrade - it only installs 14:38:46 and the new upgrade tool, fedup, isn't ready for primetime yet 14:39:22 can anyone confirm that send key->ctrl-alt-f[2-6] don't seem to work in VMs? https://bugzilla.redhat.com/show_bug.cgi?id=871241 14:40:01 eblake: How closely are you following this page: https://fedoraproject.org/wiki/Upgrading_Fedora_using_yum, and did you shut it down to runlevel 3 before doing yum distro-sync? 14:41:02 I'm wondering if it's safe to have that run while I'm using the machine as my desktop, or if I should move my xchat and mail data to another machine and setup camp there... 14:43:04 I didn't go to runlevel 3, but I am using that page as a guide 14:43:17 and I'm not using that machine for anything at the moment 14:44:02 (yes, I've set up my IRC/mail camp on a different machine today, to ensure I remain connected here no matter what happens on my test machine) 14:51:04 laine: I used runlevel 3 this time, but in the past I haven't, without any particularly bad problems. 14:52:04 crobinso: virt-manager sends ctrl-alt-f2 fine here. I posted my qemu command-line on the bug so you can see the config, it's using qxl. 14:54:19 crobinso: is it known that virt-manger hangs from randomly (ui freeze) when using qxl + spice? 14:54:30 crobinso: connecting to the vm using spicec works fine 14:54:59 drago01: haven't heard that, for f18 at least. is this a new VM or a pre-existing one? 14:55:00 s/from/f/ 14:55:34 crobinso: 3 (old) vms 14:55:46 drago01: can you pastebin one of the XML configs? 14:55:48 drago01: Are you connecting to a remote libvirtd? 14:55:50 crobinso: guests are debian, windows 7 and fedora 18 14:55:56 stefanha: no 14:56:44 there is actually a bug from f16 timeframe that causes qxl+virt-manager to hang. problem is to fix it you need to bump the qemu machine type 14:56:58 probably need to add some bit to virt-manager that can check for that 14:59:10 crobinso: oh ok let me try that 14:59:28 crobinso: is there a magic command to bump them? 14:59:36 crobinso: or should I edit the xml by hand 15:00:13 crobinso: hvm 15:00:18 crobinso: bump to what? 15:00:21 0.15? 15:00:25 drago01: virsh edit, then remove the machine= 15:00:35 just remove it? 15:00:35 ok 15:00:58 drago01: if you remove it it will choose the latest default. this might cause your windows vm to want to reactivate though, not sure if that's a problem 15:01:58 crobinso: shouldn't be ... I do have plently of (free) windows licenses 15:02:02 lets see if it hangs again 15:07:43 crobinso: no hang so far ... will report back if it breaks 15:08:06 crobinso: and yeah virt-manager should probably do that 15:08:21 the "edit the xml to fix" is not really obvious 15:16:03 crobinso: hangs again :/ 15:16:16 drago01: which guest? 15:16:22 crobinso: fedora 15:16:32 drago01: F17, F18, ? 15:16:36 crobinso: F18 15:16:46 crobinso: http://fpaste.org/OJdy/ 15:18:50 drago01: does fpaste actually work for you? it always gives me 500 internal error 15:19:02 drago01: you can use http://paste.stg.fedoraproject.org/ 15:19:02 crobinso: yes 15:19:21 crobinso: http://paste.stg.fedoraproject.org/1407/ 15:20:12 drago01: oh did you edit an /etc/libvirt/ in place? you don't want to do that 15:20:25 crobinso: yes 15:20:35 drago01: http://wiki.libvirt.org/page/FAQ#What_is_the_.27virsh_edit.27_command_and_how_do_I_use_it.3F 15:21:23 drago01: use virsh edit and restart the VM, then it will definitely use the new machine type. you can confirm your VM was using the old machine type if you pastebin /var/log/libvirt/qemu/$vmname.log 15:21:34 crobinso: oh ok 15:24:10 wtf, 'ifconfig' is not installed as a default on F18-beta-tc6 dvd :) 15:24:25 crobinso: the first time I ran virt-manager on a new machine, it said something about needing to install [something about libvirt and KVM in all caps] in order to create and run kvm guests, then asked me if I wanted to install it. When I said yes, it brought up a dialog list showing dependencies it was going to install, includeing a bunch of xen, uml, and lxc things. 15:25:05 (sorry, I didn't save the error message, as I assumed I would be able to reproduce. Unfortunately it doesn't come up on subsequent runs, even if I don't install what it asked for). 15:26:12 crobinso: what exactly is virt-manager checking for, and what package is it trying to install? Is it maybe installing the catchall "libvirt" package? If so, it should do something more limited for F18+, to prevent pulling in all that stuff that most people don't want. 15:26:32 crobinso: windows now complains about the "redhat qxl gpu" i.e driver no longer works / accepted 15:26:38 crobinso: do I need an updated one? 15:26:39 laine: yeah we basically install 'qemu-system-x86' and 'libvirt'. the 'libvirt' package pulls in all those drivers like UML, LXC etc due to the packaging back compat 15:26:56 drago01: yeah I believe so 15:27:25 These days installing "libvirt" seems a bit drastic. Should I file a BZ? 15:27:27 crobinso: pointer to it? I installed it using "spice-guest-tools-0.1" 15:29:57 drago01: hmm, not sure actually. lemme ask 15:30:22 laine: yeah BZ please 15:30:34 laine: I think so - I agree that in F18, no package should depend on libvirt directly 15:31:00 Okay. Against F18 virt-manager unless you say otherwise. 15:31:00 that the package remains for upgrade reasons and for convenience for end users, but all other packages should be picking just what they really need 15:31:18 laine: yeah virt-manager 15:32:05 ugg - my machine froze during booting again - about 50% of my boots, the kernel hangs on a line that says: 15:32:20 fb: conflicting fb hw usage nouveaufb vs VESA VGA - removing generic driver 15:32:28 power cycling usually gets it to work again 15:32:36 but I wish that the kernel wouldn't hang like that 15:33:04 don't know if it is a kernel bug, or a grub2 bug 15:33:38 and it doesn't help that the grub2 was installed back in F16 beta days - I wonder if reinstalling a newer grub2 would clean things up 15:35:02 drago01: I don't think there is anything more recent. does reinstalling the driver work? I wonder if it just got upset because the underlying hardware appeared to change 15:35:05 yay running through the I/O throttling test turned up a bug 15:39:56 crobinso: oh it complains about missing signature 15:41:35 * drago01 just disables signature checking 15:46:36 yeah something weird was going on with firefox+squid+https and fpaste.org. manually going to https://fpaste.org and accepting the cert makes things work for me now 15:48:23 Xen seems to work in my F18 beta-tc6 host 15:48:50 just yum install xen xen-hypervisor and reboot 15:49:02 let's see if libvirt/virt-manager works 15:55:21 hmm 15:55:30 installing a new PV guest with virt-manager fails with: 15:55:33 Unable to complete install: 'POST operation failed: xend_post: error from xen daemon: (xend.err "Boot loader didn't return any data!")' 15:55:49 let me try with virt-install with some debugging 15:56:09 * crobinso steps out to grab some food 16:07:23 stefanha (or anyone) what would you want to hear about at a KVM Forum talk about libguestfs? 16:22:58 gotto go (dinner, etc.) I'll try to get back here tonight 16:40:33 Tried RHEL6 with new virt-install and used libguestfs related changes on the image, seems to be working fine! 16:44:32 excellent 16:48:28 * crobinso back 17:00:12 eblake, or anyone: what's the recommended way to do freshly compiled qemu from git, w/o disturbing the system qemu pkgs? (just a 'make install' ?) 17:01:02 you mean, for pointing libvirt to use the self-built qemu but without installing qemu? 17:01:13 I think you just modify the line of your guest config 17:01:18 but haven't really tested that recently 17:01:49 right, that's what I do just change the emulator to point to the new binary .... 17:02:26 that /will/ use kvm kernel module right, (when libvirt invokes the qemu emulator) - so far it did :) 17:03:32 RHEL5 vm works fine too, with virt-install! libguestfs worked too! 17:33:44 RHEL4 vm works while installing with virt-install! libguestfs successfully placed files in the vm! 17:34:20 zer0c00l: cool! can't say I've ever installed anything older than RHEL5 17:34:57 Freebsd check coming up :) 17:35:01 zer0c00l: this is all on F18 ... what version of libguestfs did it give you? 17:35:25 and selinux on/off? (I'm guessing it must be off ...) 17:37:44 rwmjones: selinux is off 17:38:17 rwmjones: Its Fedora 17 with the test virt repos enabled 17:39:58 libguestfs-1.18.9-1.fc17.x86_64 17:41:18 it's unfortunate we don't build libguestfs in virt-preview 17:42:54 I see. Do i need F18 to test the new libguestfs? 17:47:13 yup 18:41:47 crobinso: Are we recording the results of Test Day somewhere? 18:43:08 stefanha: https://fedoraproject.org/w/index.php?title=Test_Day:2012-11-01_Virtualization#Test_Results 18:48:34 crobinso: Thanks, updated 18:53:10 crobinso: thx btw. annoying hanging issue seems solved now (no longer need to use spicec) 18:53:44 drago01: good to hear. I've got it on my list to find a way to auto-fix it 18:54:10 crobinso: ok 18:59:19 crobinso, from the live-backup test, after " mv /var/lib/libvirt/images/f18.img /my/backup/dir" you wrote : 18:59:25 If you want to get back to your original VM image name, you can now do: # virsh snapshot-create-as f18 snapshot2 "snapshot for backup" --disk-only --atomic --no-metadata --diskspec vda,file=/var/lib/libvirt/images/f18.img 18:59:40 did you mean /path/to/backup ? there, in the diskspec? 19:00:05 now that f18.img won't be in /var/lib/libvirt/images (as you've 'mv'ed it) 19:01:01 kashyap: no I think that line is correct. it's saying 'create a disk only snapshot, and make the snapshot file be f18.img' 19:01:37 ah, 'name' 19:01:38 kashyap: so if for some reason you don't like that after the backup, your VM is using a file named f18.snapshot1, you can do another set of commands and get back to a file named f18.img 19:01:49 right, you meant to change the name 19:02:23 crobinso, disregard me. That's what I get when I see wiki pages at 00:31 :) 19:02:24 yeah. I admit it's a little contrived, but I was a bit irked that the instructions didn't easily leave the VM in the config state that it started 19:02:33 :) 19:03:18 crobinso, yeah, I wrote a slightly verbose one here(reviewed by Eric), for a demo I'm giving -- http://kashyapc.fedorapeople.org/virt/lc-2012/live-backup-with-external-disk-snapshots-and-blockpull.txt 19:04:10 kashyap: yeah I looked at that, good doc! 19:04:29 kashyap: I look forward to having some graphical UI simplifying it all :) 19:05:27 crobinso, nice 19:27:36 Having a bit of a problem when running a virt-manager viewer to a nested host that is itself running virt-manager viewing a RHEL6 guest. 19:28:52 laine: how so? spice over spice just too choppy? 19:28:53 As soon as I click in the RHEL6 guest, it "captures" the mouse and tells me to press Control_L+Alt_L to release the pointer. But the outer viewer doesn't pass that in to the nested host, and 19:29:07 No actually the performance is reasonable. 19:29:27 laine: ah. yeah I'd recommend just using virt-manager on the host to connect to the F18 guest like a remote libvirt server 19:29:44 laine: so virt-manager --connect qemu+ssh://root@192.168.122.XXX/system 19:30:08 We have a freebsd disk image it works with virt-install! 19:30:26 Since it is a custom built freebsd disk image i couldn't check libguestfs 19:30:26 Right, good idea :-) My alternative was going to be ssh into the nested host and send the keystroke with virsh. 19:30:43 Tried libguestfs on it once, but it was not able to find the root file system! 19:31:04 Still would be nice to be able to passthrough an arbitrary keystroke from virt-manager's viewer though... 19:32:42 laine: yeah I hear ya, just like a 'send-key' option 19:32:49 zer0c00l: right, there's a problem with the kernel driver for BSD filesystems that stops that from working 19:33:21 https://bugzilla.redhat.com/show_bug.cgi?id=541618 19:34:04 actually the real problem is the UFS format exists in multiple variants which are not distinguished from each other by the magic in the superblock 19:35:41 And the problem of connecting in this manner is that I end up not testing F18's virt-manager (since my host isn't F18). Ah well. 19:36:05 laine: true. you could do ssh -X to the F18 VM 19:36:05 rwmjones: ok 19:36:50 "too many alternatives, entering self-protecting shield mode!" 19:37:33 :) 19:59:23 What is a blockpull? 19:59:32 i am just curious! 20:00:21 man page! 20:10:12 I have a vm that runs in runlevel 3 20:10:24 The vm was created using --nographic option 20:10:29 It is rhel 6.2 20:10:47 i did a pm-suspend from the guest, but tried to wake it up! 20:10:51 It doesn't work! 20:14:45 zer0c00l, blockpull in short, lets you flatten a snapshot image chain; by populating content from all your backing image chain into the current 'active layer' 20:16:07 kashyap: ok 20:16:24 its like replacing the entire disk blocks with one in snapshot? 20:16:38 zer0c00l, for instance, say you have this chain [base] <- [sn1] <- [active], & you want to make 'active' a self-contained, standalone image by pulling data from [base] & [sn1] 20:16:45 you could do it by issuing this command: 20:16:58 # virsh blockpull --domain f17-base --path /var/lib/libvirt/images/active.qcow2 --verbose --wait 20:17:24 (the above is essentially populating 'active' with all the contents from its entire backing chain, live) 20:20:07 ok 20:28:10 later folks! 20:33:17 Bummer, I was hoping that I could just "copy host CPU capabilities" into my nested host, and import the guests from the real host that had previously done the same thing, but some of the flags don't get copied... 20:34:23 For one thing, the "model" changes from "Operation_G3" to "qemu64". 20:37:10 Can someone try doing "Copy Host CPU Configuration" on a *real* host and see if the guest is able to start? When I do that with a WinXP guest on my nested host, it complains "internal error cannot set CPU affinity on process 0: Invalid argument" 20:40:02 laine, are you doing nested virt on amd or intel? just curious 20:40:30 (I've only tried linux guests though) 20:42:14 also you could use this to get capabilities on host 20:42:17 # virsh capabilities | virsh cpu-baseline /dev/stdin 20:47:39 kashyap: this is on AMD. 20:49:08 The hardware is actually running F16 + virt-preview, and the nested host is F18. I successfully ran a RHEL6 guest inside the F18 nested host, but now I'm trying to run a WinXP guest (which I copied directly from the hw host) and it's giving the error above. 20:56:49 okay, just rebooted into F18 (the yum upgrade appears to have worked!) - but now I'm seeing: 20:56:59 # setsebool -P virt_use_nfs on 20:57:01 setsebool --helplibsepol.print_missing_requirements: howl's global requirements were not met: bool init_systemd 20:57:02 libsemanage.semanage_link_sandbox: Link packages failed 20:57:04 Could not change policy booleans 20:57:11 huh? 20:57:22 laine, ah, I see. I'm yet to try a winxp guest though 20:57:26 how am I supposed to be able to access my disk images residing on NFS? 20:57:57 ah, I got it to work if I don't use -P, but then the setting will (probably) be lost on reboot 20:58:19 is this a bug in setsebool, or just an artifact of me doing F17->F18 upgrade instead of a fresh install? 21:01:57 eblake: not sure what's going on there, setsebool -P works fine on my machine 21:02:43 wonder if I need a selinux relabel pass? 21:03:58 at any rate, I haven't rebooted yet, but the non-persistent setting is letting me run my pre-existing guests 21:06:56 Well, I can run a WinXP guest on an F18 host that's actually a guest on an F16 host. And a macvtap device on the guest is able to communicate over an isolated bridge on the real host. But... 21:07:39 When the F18 nested host has a macvtap interface, and I try to give the nested guest a macvtap interface, for some reason it's unable to speak to the outside world. 21:08:01 I'm sure that's a *very* common use case :-P 21:10:24 BTW, the config line causing the "can't set CPU affinity" error above was this: 21:11:39 6. I haven't tried it yet, but I'm pretty sure it's because the hw host has 6 cpus, but the nested host was only given 2 :-P 21:12:49 that reinforces my idea that we should support named placements, rather than hard-coded lists, similar to how you fixed things to support named network usage 21:13:09 well, as an alternative to hard-coded lists (we can't take away existing functionality) 21:33:19 well, I used my F18 box to fire up a RHEL 5 VM, then in that VM started up a self-built libvirtd from libvirtd - and it segfaults at startup in nodeinfo.c 21:33:31 doesn't impact F18, but today's turning out to be useful 21:42:01 crobinso: eblake: everything going okay? sorry i'm sucking at checking in on test days this cycle :( 21:42:08 i threw out a quick announce on forums and twitter 21:44:59 we've at least had a couple of bugs found and fixed 21:45:57 I ran into a weird issue with setsebool, but don't know whether it is my machine (where I yum upgraded from F17 to F18 today) or something everyone would hit 21:46:00 so I haven't filed a BZ yet 21:54:13 adamw: yeah things have gone pretty good! 21:54:19 great 21:54:38 is the issue with setsebool easily reproducible? I can test it on a few f18 installs if so. 21:56:29 adamw: at least I couldn't reproduce it 21:59:51 adamw: I did 'setsebool -P virt_use_nfs on' and got a weird failure message 22:00:17 libsepol.print_missing_requirements: howl's global requirements were not met: bool init_systemd 22:00:18 libsemanage.semanage_link_sandbox: Link packages failed 22:00:20 Could not change policy booleans 22:00:33 but without -P, it changed things for the current boot, enough for me to boot VMs 22:03:30 works on my dirty desktop... 22:03:33 let's see in a clean install VM 22:04:52 works there too... 22:05:57 it may that I haven't done 'touch /.autorelabel' and rebooted - since the machine was running F17 yesterday, it could easily be an issue with my upgrade and not an impact to brand new F18 installs 22:06:08 and since I upgraded via yum instead of fedup, all bets are off 22:06:45 * eblake then again, it sounds like all bets are off with fedup as well, at least for another day or two 22:25:06 well, I have to sign off now, but I'll stay in the channel and read it later tonight 22:28:02 eblake: interesting that your RHEL5 build of libvirt git failed - I did that a few days ago and it ran without problems. 22:30:04 Also, re "named placements" I guess you mean naming the CPU cores or something? I agree with that, and also your earlier idea to have other resource limits config tied to the host rather than to the guest (I forget the exact context of that discussion too, maybe it was about max memory or something ?) 22:30:32 the build is succeeding; it's running the build that is failing 22:31:25 named placements - something like 'cpuset="seta"', where we then have an external virCgroupPtr named "seta" that points to cpus 1-3 on host A; and cpus 4-7 on host B 22:31:54 then migration between the two hosts wouldn't care that the host cpu where the guest is pinned is different according to the two hosts, just that a named set exists to control the pinning in the manner best for each host 22:31:57 eblake: i always upgrade with fedup, fwiw. the 'dirty desktop' i mentioned has been upgraded from like f15 or something, all via yum. 22:31:59 sigh 22:32:01 s/fedup/yum/ 22:34:44 adamw: Have you upgraded an F17 to F18 with yum recently? should I take the plunge and update my main desktop to F18? Or would that be too dangerous? 22:35:18 Also, do you know if a yum upgrade would work from F16 to F18? Or would I need to go to F17 first? 22:35:54 laine: i haven't tried 16 to 18. 17 to 18 with yum is actually pretty solid at this point. 22:36:05 i'm finding 18 itself fine for normal use 22:36:10 the shitshow is all anaconda-related 22:36:29 you'll want to do the upgrade with selinux in permissive, though, or else you'll hit https://bugzilla.redhat.com/show_bug.cgi?id=844167 22:36:36 not a disaster, but easier just to avoid it. 22:37:04 https://fedoraproject.org/wiki/Upgrading_Fedora_using_yum#Fedora_17_-.3E_Fedora_18 has the specific notes, though i think the thing about reinstalling polkit and systemd is old by now. 22:40:26 heh - I'm embarrassed to admit my F16 box has selionux permissive already anyway - I didn't have the time to figure out how to setup the right labelling for samba. 22:41:43 adamw: the part of that page that set my shields up was when it mentioned the possibility of "loops in dracut and never comes up". I *hate* problems like that. 22:44:06 that's old as shit too 22:44:20 i should take that out, it was just a transient bug in the repos a couple of months back 22:45:53 laine: ^^^ 23:04:35 laine: I just cleaned up the instructions and made them a bit less scary. 23:32:49 ah, I upgraded with selinux enforcing 23:32:58 maybe that explains the setsebool issue I'm seeing 23:33:31 I hit the spinning no polkit issue 23:33:45 and following the instructions to reinstall polkit helped me out today 23:33:49 so leave that in 23:33:54 adamw: ^^^ 23:41:38 eblake: i left it in, just rejigged it a bit 23:42:11 eblake: the bug basically means that any package which tries to create users or groups in its scripts as part of the upgrade will fail, aiui 23:42:36 eblake: yum history might actually be able to show you the log of the upgrade transaction and you could see what other packages were affected, and reinstall those 23:43:02 when you do the yum reinstall the creation will succeed, because now you're using f18 selinux-policy which doesn't have the problem. 23:44:36 * nirik wonders if we could get that fixed... it's anoying. 23:45:11 it's (still) being worked on 23:45:17 turned out to be a difficult one 23:53:04 sorry for being off-topic, but i figure someone here might know the answer... is there a redhat repository where i can upload a file that's too big to attach to a bugzilla? 23:59:08 jmtt: if you have a FAS account you have space on fedorapeople.org you can use for that purpose, i think 23:59:47 see https://fedoraproject.org/wiki/Infrastructure/fedorapeople.org 00:00:44 thx for the tip. i was thinking about a bugzilla.redhat.com bug, though 00:01:13 sorry for my lack of clarity. oftentimes, redhat/fedora seem synonymous to me 00:01:32 well it doesn't really matter 00:01:41 i mean, you can just upload the file to fp.o space and link to it from the bug, right? 00:02:52 if you don't have a FAS account, i'm not aware of anything rh or fedora provides which is just open upload space, no. 00:03:23 true. i could also point to my dropbox. i have a vague recollection that there was a more official space that redhat uses for customer support files, though 00:04:52 well there might be and i just don't know about it =) 00:04:56 all i know about is fedora stuff. 00:05:28 thanks for the reply 00:55:55 #endmeeting