10:45:07 <crobinso> #startmeeting virt_test_day_f20 10:45:07 <zodbot> Meeting started Tue Oct 8 10:45:07 2013 UTC. The chair is crobinso. Information about MeetBot at http://wiki.debian.org/MeetBot. 10:45:07 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic. 10:46:07 <crobinso> good morning/afternoon/evening folks :) 10:46:22 <mrezanin> crobinso: Hi Cole 10:46:44 <atodorov> crobinso: hi 10:54:38 <atodorov> crobinso: I'm looking at the ARM on x86 test case, can you point me to a doc/RTFM where these dtb files are described and what do they do 10:54:43 <atodorov> or just tell me :) 10:55:21 <atodorov> also I see vexpress-v2p-caX where X has several different values and I wonder what this is 10:56:21 <crobinso> atodorov: dtb = 'device tree binary'. device tree is a text format that describes the hardware layout of a machine. arm hardware typically isn't self describing like x86 and pci is, so you need to give the kernel a file that describes the hw layout of the arm board it is running on 10:56:47 <crobinso> atodorov: but I don't know the difference between the different vexpress variants :) I still have plenty of learning to do myself 10:59:46 <atodorov> crobinso: thanks, that explains a lot :) 11:23:14 <tjanez> After installing F20 Beta TC2, doing 'yum groupinstall virtualization', I followed the Virtualization URL Guest Install testcase 11:24:08 <tjanez> On step 5 of 5, under Advanced options I see 'No networking' and an orange exclamation mark. Is that expected? 11:24:39 <tjanez> This is 'New VM' wizard of virt-manager 11:26:05 <atodorov> tjanez: did you reboot ? 11:26:24 <atodorov> it looks like you don't have a bridge configured and the default NAT network hasn't been created as well 11:26:32 <atodorov> I think this is not expected 11:26:36 <atodorov> crobinso ^^^^ 11:26:55 <tjanez> atodorov: no, i didn't reboot 11:27:39 <laine_> libvirt's "default" network *should* be started when libvirtd is started for the first time... 11:28:03 <atodorov> tjanez: is libvirtd running then 11:28:04 <laine_> atodorov: does "virsh net-list --all" show a network named "default"? 11:28:12 <crobinso> tjanez: not expected. I did see that once during testing but I think it's entirely a virt-manager issue, what's 'sudo virsh net-list --all' show ? 11:28:15 <crobinso> :) 11:28:38 <laine_> jynx! Don't you owe me a beer now or something? 11:28:59 <tjanez> systemctl reports libvirtd.service is enabled and running 11:29:15 <crobinso> laine_: kvm forum is coming up, I'll hand you a free beer :) 11:29:29 <tjanez> "virsh net-list --all" -> the table is empty 11:29:38 <laine_> Yeah, for once there is an oppurtunity to pay up on all those beers everybody keeps promising each other :-) 11:30:10 <laine_> So maybe the network config package isn't part of the "virtualization" meta-package? 11:30:14 <atodorov> laine_: yes, networking works for me 11:30:41 <tjanez> laine_: which package are you referring to? 11:30:51 <laine_> just a sec, I'm getting the name... 11:31:04 <laine_> libvirt-daemon-config-network 11:31:27 <laine_> That used to be part of the base libvirt package, but it got split up into several components awhile back. 11:31:33 <crobinso> tjanez: make sure you run that virsh command as root, or explicitly do: virsh --connect qemu:///system net-list --all 11:31:35 <tjanez> laine_: I can confirm this package is not installed 11:32:34 <atodorov> tjanez: it is missing in the virtualization group I guess (or is not a default/mandatory package) 11:32:39 <crobinso> tjanez: yeah that missing package is the cause :) I thought groupinstall virtualization pulled that in, I'll look 11:33:30 <laine_> That wasn't mistakenly pulled out to "fix" the problems with networking when installing virtualization in a virtual machine, was it? 11:33:44 <tjanez> cronbinso: " virsh --connect qemu:///system net-list --all" also returns an empty table 11:34:23 <tjanez> ok, I'll install libvirt-daemon-config-network 11:34:29 <tjanez> and report back 11:34:40 <tjanez> should I manually restart libvirtd.service? 11:35:36 <crobinso> tjanez: yeah, manually restart libvirtd after libvirt-daemon-config-network installs 11:36:32 <tjanez> crobinso: this worked, I now have a default network 11:37:05 <tjanez> Should I file a bug on a missing Requires or will some of you fix it? 11:39:00 <crobinso> tjanez: hmm, one second 11:39:40 <crobinso> tjanez: when you first ran virt-manager, did it ask you about installing any packages ? 11:40:32 <tjanez> No 11:42:16 <crobinso> tjanez: hmm. okay, please file a bug against 'comps' that the virtualization group is missing libvirt-daemon-config-network. FYI 'comps' is the place where we track package groups 11:42:19 <atodorov> crobinso: a question: in the snapshots test case it says: Run snapshot 2, verify ..., Run snapshot 1, .... 11:42:28 <crobinso> tjanez: and cc me on the bug 11:42:37 <bitlord> same problem with network here, as tjanez said, libvirt-daemon-config-network fixed it 11:42:46 <atodorov> how do I run a snapshot ? do you mean to shut down all running snapshots and then start that one 11:43:09 <atodorov> I don't see a way to switch between the various snapshots otherwise 11:43:41 <tjanez> crobinso: do you prefer putting it in comps vs. a hard-dependency of some other package, e.g. libvirt? 11:44:08 <crobinso> atodorov: select a snapshot, and either right click->run or hit the 'play' button at the bottom. 11:44:20 <atodorov> ok, got it 11:44:25 <crobinso> tjanez: the packages were originally split for various valid reasons so we can't mandate it 11:44:53 <tjanez> crobinso: Ok, thanks for the explanation 11:45:01 <crobinso> tjanez: for example gnome-boxes doesn't care about the network, and it can cause issues if used within a virtual machine. but for the virtualization group package set it should be included at least 11:45:56 <tjanez> crobinso: Aha, I see 11:47:17 <crobinso> atodorov: the UI might not make it clear, but think of the snapshot as a 'screenshot' of your vm. when you take a snapshot, it just kinda sits there in a file and your VM moves on. but you can 'run' that old snapshot and it will revert your VM to the that old memory and disk state. so there isn't really a concept of a 'currently running' snapshot, or more than one 'running' snapshot, they are all just points in time 11:47:25 <laine> Ugh. Many of the fedoraproject mirrors don't have the F20 Alpha iso images :-/ I had to retry several times 11:47:55 <crobinso> laine: yeah the f20 mirrors seem to be in rough shape, I've hit that too 11:48:15 <atodorov> crobinso: I got confused by the State: Running and the active icons in the UI 11:48:27 <atodorov> I guess it can be improved a bit, will file a RFE 11:48:29 <crobinso> laine: dl.fedoraproject.org is the master mirror 11:48:48 <laine> would be nice if the download page would cycle between the different mirrors until it finds one that works, like yum does... 11:49:06 <crobinso> atodorov: ah that makes sense, please file a bug about that, I can reword things 11:49:35 <crobinso> laine: yeah I suspect there's a bug somewhere, I'll ask 11:50:11 <tjanez> laine: atodorov suggested I install Beta TC2 rather than Alpha 11:50:19 <tjanez> laine: http://dl.fedoraproject.org/pub/alt/stage/20-Beta-TC2/ 11:50:32 <laine> Oh, thanks for the info. I'll stop my downloads now! 12:00:19 <tjanez> crobinso: Filled the bug and added you on Cc: https://bugzilla.redhat.com/show_bug.cgi?id=1016606 12:00:34 <tjanez> OT: Bugzilla is awfully slow... 12:00:57 <atodorov> crobinso: Launch virt-manager. Start with a shutoff VM. Verify that the guest has a 'Channel qemu-ga' device listed in its hardware details. 12:01:19 <atodorov> crobinso: is it expected this device to be there by default or not? I don't see it but I can add it and start the test 12:02:30 <crobinso> atodorov: it should be there by default. one or yours bug indicated your were missing virtio console as well. what virt-manager version do you have? 12:03:22 <crobinso> atodorov: but yes, please add it and continue with the test. should be able to do 'add hardware'->channel->select the org.qemu name and it should set up the correct default 12:03:40 <crobinso> tjanez: thanks for filing that 12:03:54 <atodorov> crobinso: I'm using virt-manager-0.10.0-4.git79196cdf.fc20.noarch 12:04:05 <atodorov> I will file a bug for the missing hardware then 12:04:28 <crobinso> atodorov: did you definitely specify fedora20 for your VM OS? (in the 'new vm' wizard) 12:04:53 <crobinso> atodorov: or fpaste your ~/.cache/virt-manager.log and I can confirm 12:04:59 <atodorov> I was using URL to install from and I think it was correctly recognized as such 12:05:24 <crobinso> atodorov: hmm, okay, please file a bug and attach your virt-manager.log 12:06:49 <atodorov> crobinso: ok 12:24:16 <crobinso> eblake: mornin 12:24:23 <eblake> hello! 12:24:48 <eblake> which is better? run tests under F19 + fedora-virt-preview, or upgrade my bare-metal machine to full F20? 12:25:51 <atodorov> mrezanina: ping, how long did you have to wait for the arm guest to start ? nothing happens for me, just black screen. Also did you use the +lpae kernel/initrd or not ? 12:28:20 <tjanez> eblake: If you'll do a fresh install, atodorov suggested using Beta TC2 12:28:30 <tjanez> eblake: I don't know about the upgrade 12:29:01 <atodorov> crobinso: ^^^^, I've followed the ARM test case instructions but nothing happens and I'm kind of lost here 12:29:19 <eblake> tjanez: I'd rather do a fedup or yum upgrade to my F19 partition; but it is a quad boot machine, so I guess I could wipe my existing rawhide partition (which hasn't booted lately anyways) and replace it with a fresh TC2 install 12:30:03 <crobinso> eblake: I updated to f20 and it's been pretty smooth so far 12:30:40 <crobinso> atodorov: don't use lpae kernel, it didn't work for me :/ 12:30:42 <eblake> crobinso: with fedup, or just yum? 12:30:47 <crobinso> eblake: yum 12:31:15 <atodorov> crobinso: ok, will try the other 12:31:19 <crobinso> atodorov: but you should start seeing output within 5 seconds, if not then it's configured wrong 12:31:49 <crobinso> atodorov: this stuff is pretty touchy, and the failure scenarios suck but qemu devs know about it's not a priority 12:32:04 <crobinso> *know about it and it is not a priority 12:35:17 <tjanez_> I abandoned the 'Guest install via URL' test case due to mirror problems for F20 Alpha 12:35:57 <tjanez_> Now I tried to use the Beta TC2 I downloaded and I'm stuck with the following virt-manager error: http://ur1.ca/fuv68 12:37:12 <tjanez_> the permissions of the ISO image are the following: -rw-rw-r--. qemu qemu system_u:object_r:virt_content_t:s0 Fedora-Live-Desktop-x86_64-20-Beta-TC2.iso 12:38:11 <atodorov> crobinso: non pae kernel+initrd moved past the boot point, but now I'm stuck with growroot Fatal: Failed to re-mount /dev/vda3, this is bad 12:39:01 <atodorov> oops, priot to that it says: mount: special device /dev/vda3 does not exist and priot to that it shows sfdisk -d which clearly shows vda1, 2, 3 and 4 12:39:31 <crobinso> atodorov: can you show the guest XML ? 12:41:04 <crobinso> tjanez_: hmm, did virt-manager ask you anything about changing your home directory permissions ? 12:41:51 <atodorov> crobinso: http://www.fpaste.org/45073/ 12:42:00 <tjanez_> crobinso: No. I've given the output of 12:42:23 <tjanez_> crobinso: 'ls -alZ' and it looks ok to me 12:43:24 <crobinso> tjanez_: it's more subtle then that: the qemu user need search permissions all the way down to the iso file. can you file a virt-manager bug and attach ~/.cache/virt-manager/virt-manager.log, and that error message 12:44:01 <crobinso> tjanez_: the workaround is move the iso to /var/lib/libvirt/images, restart virt-manager, and continue. that directory already has the correct permissions and labeling 12:44:12 <crobinso> tjanez_: so maybe kick off the install, _then_ file the bug :) 12:44:26 <tjanez_> crobinso: Ok, thanks, will do that :) 12:45:22 <atodorov> tjanez_: just setenforce 0 if you have your files in non standard location 12:46:08 <tjanez_> atodorov: well, I could do that, but wouldn't testing with SELinux enabled be of more value? 12:47:16 <crobinso> tjanez_: also in that bug report ls -ld $HOME and ls -ld $HOME/Downloads, and getfacl for each 12:47:22 <atodorov> tjanez_: yes, but that depends on where your focus is. right now I prefer giving virt-manager a spin without selinux getting in the way 12:48:28 <crobinso> tjanez_: I would be interested as well if setenforce 0 makes that case work, since that limits it 12:48:43 <crobinso> atodorov: I can't tell what the issue is from your XML, looks fine there :/ 12:48:53 <crobinso> atodorov: this is using the beta images? 12:49:11 <atodorov> crobinso: yes, using Beta-TC2, arm with MATE desktop 12:49:17 <atodorov> should I try the minimal one 12:49:36 <tjanez_> atodorov: I don't have a special focus, I was just asking for your (and other developers') opinion :) 12:49:52 <atodorov> and can I use the same vmlinuz+initrd without extracting them again ? 12:50:14 <crobinso> atodorov: not sure 12:50:23 <atodorov> crobinso: ok, will try 12:50:28 <crobinso> atodorov: I'll try and reproduce with the beta images, I only tried alpha and minimal myself 12:50:51 <atodorov> crobinso: btw I think I know the answer to my earlier question about the a9, a15 dtb images 12:50:56 <crobinso> tjanez_: I run with setenforce 0 but I watch for selinux errors reported by sealert 12:51:26 <atodorov> together with cpu arch == arm you have to select the CPU type in virt-manager and this is where these variants come into play 12:52:23 <crobinso> atodorov: right I know they largely map to different boards/machine types. but there's like 3 vexpress-a9 dtb and I don't know the difference between those :) 12:54:25 <tjanez_> atodorov, crobinso: I just did 'sudo setenforce 0' and re-clicked the 'Finish' button on Step 5 of 5, but it still returns the same error 12:54:45 <crobinso> tjanez_: okay, please include that info in the eventual bug report 12:55:25 <tjanez_> crobinso: Ok, will do. I'll move the image to /var/lib/libvirt/images now... 13:05:13 <kashyap> eblake, Yes, I can attest to it too, F20 update went just fine: yum update yum; yum clean all; yum --releasever=20 distro-sync --nogpgcheck -y 13:13:25 <crobinso> atodorov: downloading the beta image now, so i'll try to confirm in a bit. 13:15:00 <atodorov> crobinso: tested with the TC2 minimal image, same isssue. filing a bug right now 13:15:36 <crobinso> atodorov: sounds good, please include guest XML, and /var/log/libvirt/qemu/$vmname.log 13:22:05 <atodorov> crobinso: 1016648 13:42:13 <roshi> crobinso: how's the test day going so far? 13:43:27 <crobinso> roshi: we've generated a lot of bug reports. so, successful :) 13:44:43 <roshi> good to hear! 13:49:32 <atodorov> roshi: 8 bugs from me :) 13:49:38 <atodorov> 2 for anaconda though 13:49:49 <roshi> awesome atodorov 13:49:52 <roshi> good work 13:50:17 * roshi just woke up 14:10:59 <crobinso> atodorov: hmm, qemu-system-arm crashes for me with the the beta :/ but I don't think I'm running the latest qemu build 14:11:00 <dustymabe> hey guys does anyone get random guests that are left around after creating a guest? 14:11:05 <dustymabe> 25 guestfs-f8z0ja20o9ys8zs8 paused 14:11:43 <atodorov> dustymabe: where do you see this ? 14:11:46 <crobinso> dustymabe: those are libguestfs temporary guests... though they shouldn't be paused I don't think. 14:12:16 <atodorov> crobinso: I had seen the VM guest and sometimes anaconda not shutdown gracefully but I'm not able to reproduce consistently 14:12:19 <atodorov> can this be related 14:12:25 <dustymabe> basically after I do an install 14:12:34 <dustymabe> I'll work on getting a solid reproducer 14:12:56 <atodorov> dustymabe: that could be a the anaconda issue I'm talking about ^^^ 14:13:40 <crobinso> dustymabe: ah, I bet it's due to your usage of nested virt. if the guest hits a kvm error, it can go into the paused state. libguestfs launches with some equivalent of -cpu host which might be triggering kvm issues due to all the nesting 14:14:52 <dustymabe> ok.. anything you want me to do to investigate? 14:15:28 <crobinso> dustymabe: I pinged rjones, if he's around he might have ideas. /var/log/libvirt/vmname/guestfs-f8z0ja20o9ys8zs8.log might give a hint 14:16:22 <atodorov> crobinso: a question about Pause/Resume UX - I don't see the guest console becoming gray or looking like it is disabled, should it hint the user visually as to what is the current state ? 14:18:50 <rwmjones> dustymabe: did you run libguestfs-test-tool? can you paste the complete output somewhere 14:18:56 <crobinso> atodorov: we used to do that a long time ago, but dropped it because it was a pain to get right with the console scaling. it's a reasonable idea but I don't see it being implemented anytime soon :) 14:19:47 <dustymabe> rwmjones: no I have just been using virt-install 14:19:58 <atodorov> crobinso: I will update the test case then to indicate the UI is not updated 14:20:10 <dustymabe> rwmjones: the log did show me an error: KVM: entry failed, hardware error 0x7 14:20:39 <dustymabe> rwmjones: I am using host passthrough with nested virt so maybe that is why it is having trouble 14:20:39 <rwmjones> dustymabe: this was a libguestfs problem or a libvirt problem? I didn't see previous discussion 14:21:12 <dustymabe> rwmjones: I just notice on my system that I have a random guest that hangs around: 25 guestfs-f8z0ja20o9ys8zs8 paused 14:21:41 <dustymabe> rwmjones: this is after running a couple of virt-install tests.. nothing special 14:22:06 <rwmjones> hmm strange, are you running virt-manager? 14:22:48 <dustymabe> i have ran virt-manager but only to view console of the guests.. to create the guests I have been using virt-install 14:23:22 <dustymabe> Would you like for me to send you a link to the full output in the log? 14:23:24 <rwmjones> don't know, but you can virsh destroy that guest .. it shouldn't be there and it should be in paused state 14:24:11 <rwmjones> s/should be/should NOT be/ 14:24:37 <dustymabe> here is a link to the log file: http://pastebin.com/xgU6WgB2 14:25:18 <rwmjones> looks badly broken; could be nested KVM or something else KVM-related 14:26:07 <dustymabe> ok.. I'll keep it in mind and continue testing 14:26:58 <dustymabe> i do have one other nested virt question.. has anyone else seen NMI's in L2 guest? Similar to this: https://bugzilla.kernel.org/show_bug.cgi?id=58941 14:27:50 <dustymabe> The guest continues to run fine but the messages get printed to console periodically 14:47:07 <dustymabe> is it still valid to create a guest with --graphics none and expect to see anaconda install on the serial console? 14:47:17 <crobinso> dustymabe: haven't heard of that NMI issue, but not surprising :) nested vmx is finicky 14:48:29 <crobinso> dustymabe: did anaconda ever do it automatically? I though passing in a kickstart, or at least 'console=ttyS0' was required 14:48:39 <crobinso> to get the non-graphical install 14:49:50 <crobinso> but if you're doing --os-variant fedora20 you'd likely need to tell it 'console=hvc0' since that's the virtio console inside the guest, but I haven't confirmed that actually works 14:50:13 <dustymabe> If i recall correctly, at least in RHEL6 if you passed in --nographics it would default to connect you to the serial console where you could receive messages from ttyS0 14:50:21 <dustymabe> and I passed in console=ttyS0 14:50:23 <dustymabe> ahh got it 14:50:29 <dustymabe> so it is not ttyS0 anymore? 14:50:58 <dustymabe> i'll try with console=hvc0 14:55:48 <dustymabe> crobinso: bingo.. i'm now getting messages on the serial console 14:56:04 <atodorov> crobinso, dustymabe: https://bugzilla.redhat.com/show_bug.cgi?id=1016715 , FYI. I filed the poweroff hang issue which unfortunately is not easy to reproduce :( 14:56:16 <atodorov> if you see it again please add more info to this bug 14:56:31 <atodorov> I have to go now. enough testing for me today 14:56:45 <crobinso> atodorov: thanks for all the help! 14:56:55 <dustymabe> atodorov: were you using nested virt? 14:57:36 <atodorov> dustymabe: nope, just bare metal F20 14:58:05 <dustymabe> atodorov: thanks.. just trying to narrow down if nested virt was the reason for what I saw or not. If i see it again I'll try to narrow it down 14:58:23 <crobinso> dustymabe: yeah I changed the default for f20 to virtio console, but I was afraid it would causes issues like that. 14:59:20 <crobinso> dustymabe: virtio console is nice because after install, systemd knows to start a getty on it, so you can do: virsh start --console $vmname and get a login prompt with zero extra config 14:59:48 <crobinso> dustymabe: but for people doing headless installs it means passing different kernel args... 14:59:49 <dustymabe> crobinso: with "--graphics none" could virt-install append console=<appropriate console> to the kernel command line? 15:01:24 <dustymabe> crobinso: I think systemd will do that for any serial console that is provided to the kernel command line (not just hvc0) 15:01:51 <crobinso> dustymabe: yeah we could add console automatically, or warn if it's likely not going to work 15:02:26 <crobinso> dustymabe: right systemd does that automatically if the grub kernel command line is altered. but the nice thing about hvc0 is that systemd does it _regardless_ of the kernel command line, so zero extra config is needed 15:02:50 <dustymabe> crobinso: got it 15:03:28 <dustymabe> crobinso: how do you track feature requests such as the one mentioned above? 15:04:04 <crobinso> dustymabe: just file a fedora virt-manager bug and I'll move it around if necc. 15:05:15 <dustymabe> crobinso: got it.. this will help lessen your worries about switching the default console (at least for initial install) 15:06:15 <dustymabe> ok.. next on the list 15:06:45 <dustymabe> i created a guest without graphics.. and then later attempted to add spice graphics to the VM using virt-manager 15:07:20 <dustymabe> I get the following error: 15:07:36 <dustymabe> details=Error adding device: XML error: Attempted double use of PCI slot 0000:00:02.0 (may need "multifunction='on'" for device on function 0) 15:08:07 <dustymabe> this is because the network card was already using that pci slot 15:09:08 <crobinso> dustymabe: hmm, can you pastebin ~/.cache/virt-manager/virt-manager.log ? 15:09:41 <crobinso> dustymabe: and the full guest XML 15:10:44 <dustymabe> crobinso: yes. let me truncate the log and recreate real quick so there are only messages we care about 15:10:54 <crobinso> thanks 15:17:34 <dustymabe> crobinso: Here is log: http://pastebin.com/zjNd2FUq 15:19:29 <dustymabe> crobinso: Here is the guest xml http://pastebin.com/jSbfs7f7 15:19:30 <crobinso> dustymabe: I'll check it in a bit, I'm on a phone call now. did you work around the issue in the meantime? 15:20:00 <dustymabe> crobinso: yep.. i just wanted to bring it up 15:20:34 <crobinso> dustymabe: sounds good. please file virt-manager bug about that and I'll follow up there 15:21:32 <dustymabe> ok 15:25:30 <dustymabe> tjanez_: are you still around? I am also having issues with installing via a URL.. Is anyone else? 15:25:33 <dustymabe> Warning: Downloading 'http://download.fedoraproject.org/pub/fedora/linux/development/20/x86_64/os//images/product.img' failed! 15:26:33 <crobinso> dustymabe: yeah mirror manager is messed up, replace download.fedoraproject.org with dl.fedoraproject.org and it should work 15:27:35 <crobinso> download.fedoraproject.org will automagically choose a mirror, but some of the mirrors are incomplete so bits will fail. dl.fedoraproject.org is the master fedora mirror 15:27:45 <dustymabe> ok thanks 15:48:47 <tjanez_> dustymabe: As crobinso pointed out, the state of development mirrors (or F20 Alpha mirrors) is quite bad. Anaconda even crashed on me during install due to a missing rpm. 15:48:59 <tjanez_> dustymabe: I just downloaded an ISO and used that. 15:49:49 <dustymabe> tjanez_: thanks.. i am using dl.fedoraproject.org now.. seems better 15:50:13 <tjanez_> crobinso: I'll have to leave now, is it still valuable to perform more test cases later or tomorrow? 16:03:34 <crobinso> tjanez_: yes please, any time this week is fine really 16:03:40 <crobinso> tjanez_: thanks for the help :) 16:04:49 <tjanez_> crobinso: No problem, I was glad to help out :) Will do some more tests later. 16:19:14 <crobinso> dustymabe: file those bugs yet? I've got some info to dump :) 16:20:33 <dustymabe> crobinso: not yet will do that now.. somehow my whole system went down at the end of an network install.. i couldn't switch to a different vty or anything 16:20:46 <dustymabe> working on the bug reports now 16:20:56 <dustymabe> i'll file the virt-manager one first 16:21:22 <crobinso> ugh, system crash. no good. 16:35:03 <dustymabe> crobinso: https://bugzilla.redhat.com/show_bug.cgi?id=1016775 16:38:51 <kashyap> dustymabe, I was just reading some scroll, if you're using nested virt (kvm on kvm), please ensure you're having versions: 16:38:52 <kashyap> 3.12.0-0.rc3.git1.2.fc21.x86_64 16:38:53 <kashyap> qemu-kvm-1.4.2-7.fc19.x86_64 16:38:53 <kashyap> libvirt-daemon-kvm-1.0.5.5-1.fc19.x86_64 16:39:00 <kashyap> (Or higher) 16:39:57 <crobinso> kashyap: does yum --enablerepo=rawhide update kernel 'just work' or does it want to pull in the world ? 16:39:58 <kashyap> Here is the link for No debug (i.e. slightly faster) Rawhide kernels -- http://fedoraproject.org/wiki/RawhideKernelNodebug 16:40:37 <kashyap> crobinso, Good question, it should be fedora-rawhide (exact repo name), no? 16:40:43 <kashyap> Btw, good morning! 16:41:40 <crobinso> kashyap: good evening to you :) 16:41:41 <kashyap> crobinso, Ok, that doesn't work :-) 16:41:48 <dustymabe> kashyap: thanks for the info. my kernel is older (from F19) 16:41:48 <kashyap> I just use the nodebug kernels 16:42:01 <kashyap> crobinso, Rawhide kernels have debug enabled. 16:42:35 <kashyap> crobinso, Yeah, I'm about to wind up and take a walk :-) 16:46:27 <dustymabe> cronbinso: should i put the bug i filed on the test day page somewhere? where should it go? 16:46:49 <dustymabe> oops.. crobinso 16:47:09 <crobinso> dustymabe: sure, just add a line under the 'known bugs' section 16:47:15 <crobinso> dustymabe: auto complete is your friend :) 16:47:31 <kashyap> 'cron'binso - :-) 16:48:20 <kashyap> dustymabe, And, yes (to your previous question) - NMIs in L2 is not uncommon. (One of the KVM unit tests specific to NMI was failing, there was a discussion upstream for a fix, I think it should be fixed if you're trying kvm.git next.) 16:48:58 <crobinso> kashyap: so is 3.11 much better than 3.10 for nested vmx? 16:49:33 <kashyap> crobinso, Certainly. Even with 3.13.0, I had some host crashes, debugging w/ Paolo/gleb. 16:50:01 <kashyap> So, I was just installing kernels from kvm.git queue. 16:51:08 <kashyap> crobinso, Side note: With nested EPT (EPT enabled on both L0 & L1), you see 10x perf improvement compared to shadow on shadow (EPT disabled on both L0 & L1) 16:52:00 <kashyap> crobinso, By 10x, a simple example: with nEPT -- Kernel compile in L2 takes 5 minutes; with shadow-on-shadow, same Kernel compile on L2 takes 50 minutes. 16:52:36 <crobinso> nice 16:53:06 <kashyap> defconfig that is. Standard distro bloat Kernel takes more time, as it enables every darn thing :) 16:54:03 <kashyap> Anyway, stepping out for now. See ya. Will read scroll later. 16:54:19 <dustymabe> kashyap: thanks for the info :) 17:10:24 <dustymabe> Is it possible in virt-install to append kernel args to an install based off a cd? 17:10:56 <dustymabe> if you specify a url location then virt-install will detect kernel, initrd, etc.. and then feed that info to the qemu command 17:11:30 <dustymabe> if you do a cd it doesn't so there is no kernel option in the qemu command and thus the qemu command will reject any --append statements 17:11:32 <crobinso> dustymabe: nope, there's no easy way to get the kernel args to the kernel that's booted from the iso 17:12:38 <crobinso> dustymabe: it's kind of crazy, but what you can do is use virt-manager on the host machine to connect over ssh to libvirt inside the VM, then kick off a graphical install 17:13:17 <dustymabe> crobinso: yep and add your kernel command line options there 17:13:38 <crobinso> dustymabe: I though virt-install gave a warning that --kernel-args won't work for CDROM boot... did it not? 17:14:52 <dustymabe> crobinso: I don't think so.. or at least I did not notice until qemu cursed at me 17:15:17 <dustymabe> crobinso: FYI I was using --location /dev/sr0 and --extra-args for the kernel args 17:15:46 <crobinso> dustymabe: well if qemu gives an explicit error that's good enough for me 17:17:09 <dustymabe> crobinso: it wasn't immediately obvious but I did figure it out. wasn't immediately obvious because I didn't know that for the web install it was passing kernel and for the other one it wasn't 17:17:17 <dustymabe> crobinso: i figured it out though and it makes sense now 17:17:51 <crobinso> dustymabe: yeah, using --location with a cdrom kind of works and kind of doesn't, I should clarify it one of these days 17:18:12 <dustymabe> crobinso: would it make sense to do the same "detection" for cd/iso installs filtered through the --location argument as is done for URL installs? 17:18:55 <crobinso> dustymabe: I would like to, but it's not as easy :) libosinfo will help with that but we need to port to it 17:20:30 <dustymabe> crobinso: got it.. im sure you've wanted to append kernel args to cd installs too for a while 17:21:06 <dustymabe> crobinso: in an automated fashion that is 19:51:04 <atodorov> hi guys, if anyone is interested, here's my test day post mortem review - http://atodorov.org/blog/2013/10/08/fedora-20-virtualization-test-day-post-mortem/ 19:53:32 <crobinso> atodorov: great, thanks 19:57:49 <dustymabe> atodorov: was it you that opened the bug about anaconda not restarting cleanly at the end of install? 19:59:00 <atodorov> dustymabe: yes, one sec 19:59:21 <atodorov> https://bugzilla.redhat.com/buglist.cgi?bug_id=1016435%2C1016449%2C1016488%2C1016530%2C1016604%2C1016613%2C1016648%2C1016663%2C1016704%2C1016715&list_id=1785007 19:59:47 <atodorov> here's all my bugs for today, the one you are asking about is https://bugzilla.redhat.com/show_bug.cgi?id=1016715 20:01:45 <dustymabe> atodorov: thanks 20:20:26 <dustymabe> ok guys im heading out. 20:20:42 <dustymabe> crobinso: on our discussion earlier about injecting kernel args for cd install 20:21:18 <dustymabe> crobinso: I think we can get what we want by mounting the iso/cd and then pointing --location to the directory 20:21:39 <dustymabe> crobinso: then kernel args work 21:08:57 <crobinso> dustymabe: yeah, the only problem with that is then getting the install media into the guest, but you could export the mount point over http 21:46:46 <dustymabe> crobinso: you know i was just brainstorming this 21:47:51 <dustymabe> would be cool if we could share the directory to the guest using http://www.linux-kvm.org/page/9p_virtio 21:48:25 <dustymabe> although i'm sure that would be a nightmare for situations where the guest is remote from the system you are running virt-install on 21:49:06 <crobinso> dustymabe: yeah that is definitely an option, but only for sufficiently new guests. it's also not perfect: need to be root to mount an iso, and many isos are live media which aren't typical install trees, etc. 21:51:37 <dustymabe> crobinso: didn't think about the "root" part 00:55:16 <crobinso> #endmeeting