| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-11-05 | |||
| 23:27:30 | sean-k-mooney | sometimes i forget how "fun" C is | |
| 23:34:26 | efried | sean-k-mooney: how do I check my version of qemu? | |
| 23:35:00 | efried | n-cpu failed to start with | |
| 23:35:00 | efried | Nova requires QEMU version 2.8.0 or greater. | |
| 23:35:00 | efried | but the plugin should have set up 4.1.0, no? | |
| 23:35:35 | sean-k-mooney | correct | |
| 23:35:41 | sean-k-mooney | http://paste.openstack.org/show/785829/ | |
| 23:35:50 | sean-k-mooney | i have 4.1 | |
| 23:39:31 | openstackgerrit | melanie witt proposed openstack/nova stable/stein: Add regression test for bug 1824435 https://review.opendev.org/692906 | |
| 23:39:31 | openstack | bug 1824435 in OpenStack Compute (nova) stein "fill_virtual_interface_list migration fails on second attempt" [Medium,In progress] https://launchpad.net/bugs/1824435 - Assigned to melanie witt (melwitt) | |
| 23:39:31 | openstackgerrit | melanie witt proposed openstack/nova stable/stein: Remove redundant call to get/create default security group https://review.opendev.org/692907 | |
| 23:45:43 | sean-k-mooney | efried: ill try running this all again on a clean ububut cloud image in the morning and ill provide you with the local.conf and anychagnes that are needed on the system if any in an update to the readme. | |
| 23:46:17 | sean-k-mooney | one way or another its in my best interest to get this documented and updated in the repo because i will need this again at somepoint in the futrue | |
| 23:53:16 | efried | sean-k-mooney: I gotta run and break some arms, but when I run the qemu commands I see 4.1, but obv nova thinks I've got something lesser | |
| 23:53:27 | efried | anyway, thanks for the help so far, I'll bug you again tomorrow. | |
| 23:54:32 | sean-k-mooney | no worries sorry it didnt "just work" | |
| 23:55:02 | efried | :) | |
| #openstack-nova - 2019-11-06 | |||
| 02:37:40 | openstackgerrit | Archit Modi proposed openstack/nova master: Revert "openstack server create" to "nova boot" in nova docs https://review.opendev.org/693065 | |
| 02:40:19 | gibi_ptg | the nova ptg table: | |
| 02:40:21 | gibi_ptg | http://vogon.no-ip.org/gibizer/ptg/signal-attachment-2019-11-06-103414_001.jpeg | |
| 02:40:26 | gibi_ptg | http://vogon.no-ip.org/gibizer/ptg/signal-attachment-2019-11-06-103414_002.jpeg | |
| 02:40:42 | gibi_ptg | so far we can fit | |
| 04:28:31 | eandersson | We just deployed the latest nova patches for Rocky | |
| 04:28:34 | eandersson | and are getting | |
| 04:28:36 | eandersson | > Could not retrieve compute node resource provider XXX and therefore unable to error out any instances stuck in BUILDING state. | |
| 04:28:41 | eandersson | On hundreds of nodes on startup | |
| 04:36:07 | eandersson | pretty sure what happens is that in Rocky with latest stable it will error if you restart a compute node with no instances | |
| 04:36:26 | eandersson | https://github.com/openstack/nova/commit/13bb7ed701121955ba015103c2e44429927e78d4#diff-77f9348ab09642ba46409b6828af4af0R1327 | |
| 04:44:25 | eandersson | actually it happens on non-empty computes as well | |
| 05:07:34 | eandersson | Interesting | |
| 05:07:46 | eandersson | An error'd compute due to quota still takes up an allocation | |
| 05:08:24 | eandersson | Preventing heal_allocation from progressing | |
| 05:15:47 | eandersson | Turns out placement isn't in a great place in that region. Probably why we are hitting errors. | |
| 05:39:50 | eandersson | I wish heal_allocations would have an option to skip failed computes. | |
| 05:40:25 | eandersson | When you are in a bad state you basically need to do some mysql magic to figure out "bad" hosts. Otherwise you are fixing them sequentially. | |
| 08:57:03 | kashyap | gibi_ptg: Good luck :-) | |
| 10:36:12 | openstackgerrit | Boris Bobrov proposed openstack/nova master: Also enable iommu for virtio controllers and video in libvirt https://review.opendev.org/684825 | |
| 10:36:13 | openstackgerrit | Boris Bobrov proposed openstack/nova master: Create a controller for qga when SEV is used https://review.opendev.org/693072 | |
| 12:23:44 | sean-k-mooney | ... so the random number generator has been broken in nova for a while... | |
| 12:24:09 | sean-k-mooney | and by a while i mean since https://github.com/openstack/nova/commit/be548a3396ee5f3a78b2b8b59321eb4c4530f7fb merged in queens | |
| 12:25:31 | sean-k-mooney | the comments is in the wrong place | |
| 12:25:33 | sean-k-mooney | https://github.com/openstack/nova/commit/be548a3396ee5f3a78b2b8b59321eb4c4530f7fb#diff-f4019782d93a196a0d026479e6aa61b1L4037-L4071 | |
| 12:25:57 | sean-k-mooney | after teh refactor the rng is only enabled if you also enable the qemu guest agent | |
| 12:26:39 | sean-k-mooney | before it was eneabled if you enabled it in the image and the host was qemu or kvm | |
| 12:27:52 | sean-k-mooney | kashyap: ^ we proably need to fix that but we talked before about enableing the rng by default. it should be two seperate patches but do you think we shoudl still enable it by default | |
| 12:38:03 | sean-k-mooney | i have filed a bug for it here https://bugs.launchpad.net/nova/+bug/1851497 | |
| 12:38:03 | openstack | Launchpad bug 1851497 in OpenStack Compute (nova) "rng is only enabled if qemu guest agent is enabled" [Medium,Triaged] - Assigned to sean mooney (sean-k-mooney) | |
| 12:54:46 | sean-k-mooney | efried: hi so http://paste.openstack.org/show/785846/ seams to now work. im going to kick it off on a clean vm to test but that stacked and i could boot vms | |
| 13:16:13 | efried | sean-k-mooney: cool beans | |
| 13:21:39 | efried | sean-k-mooney: I just need to update that enable_plugin line for my purposes, right? The rest ought to be unchanged from yesterday? | |
| 13:22:12 | sean-k-mooney | yep | |
| 13:22:41 | sean-k-mooney | ill proably merge the patch today or tomorow but i want to do a clean install first to make sure i have not missed something | |
| 13:23:18 | sean-k-mooney | i need to actully set up a gate job to actully test it rather then manually testing it every time i update it | |
| 13:23:38 | efried | okey. Were you able to repro the last problem I had yesterday where nova thought I didn't have a recent enough qemu? | |
| 13:23:55 | efried | like, it was looking in the wrong place or something | |
| 13:24:23 | sean-k-mooney | not exactly but i noticed i was not unistalling qemu from the package manager before installing it for source | |
| 13:24:27 | sean-k-mooney | so i fixed that | |
| 13:24:33 | sean-k-mooney | i had that for libvirt but not qemu | |
| 13:25:06 | efried | ah | |
| 13:25:12 | efried | sounds like that would probably do it. | |
| 13:25:52 | efried | k, I'm running now. Cleaned first to get rid of compiled libvirt/qemu -- I needed that because of new compile flaggage, rite? | |
| 13:26:01 | efried | for apparmor | |
| 13:26:43 | sean-k-mooney | actully no i have not added that yet | |
| 13:27:07 | sean-k-mooney | but i found the bits i needed to set to disable the use of polkit | |
| 13:27:30 | sean-k-mooney | a clean does not hurt it just will be slower | |
| 13:27:35 | efried | ight. We'll see how this goes :) | |
| 13:29:04 | sean-k-mooney | by the way to go back to normal you just need to unstack then comment out the pluging. it should remove the libvirt and qemu it installed on unstack but wont delete the source in /opt/stack | |
| 13:29:19 | efried | cool | |
| 13:29:23 | efried | I'll never go back to normal | |
| 13:29:26 | efried | This is my new normal | |
| 13:29:41 | sean-k-mooney | :) | |
| 14:00:40 | efried | sean-k-mooney: no idea why this should be failing differently on my otherwise-identical VMs, but one is hitting the libvirt-python install failure again, and the other is hitting the /var/run/libvirt/libvirt-sock ENOENT again. | |
| 14:12:02 | sean-k-mooney | so the libvirt python failure is complaingin about the libvit.so right with pkg-config or somthing like that and the other case libvirt is not running? | |
| 14:13:12 | efried | sean-k-mooney: Maybe I'm not looking at the right patch set. I don't see a delta where you're purging qemu for instance. | |
| 14:14:17 | sean-k-mooney | its here https://review.opendev.org/#/c/693094/3/devstack/libs/libvirt@102 | |
| 14:14:30 | efried | ah, k | |
| 14:15:32 | efried | well, now I got both to fail the same way. The one that was failing with ENOENT is now failing on the libvirt-python thing. | |
| 14:15:52 | efried | I cut it over to USE_PYTHON3 | |
| 14:16:28 | sean-k-mooney | ok. am if you look in /usr/lib is there a libvirt.so | |
| 14:17:05 | efried | yes | |
| 14:17:20 | sean-k-mooney | ok if you unstack it should remove it | |
| 14:17:55 | efried | and that's what I want? | |
| 14:17:58 | sean-k-mooney | if it does not run " sudo rm -f /urs/lib/libvirt* | |
| 14:18:21 | sean-k-mooney | well we want to make sure its the libvirt.so you compiled for a start | |
| 14:18:28 | sean-k-mooney | it shoudl be | |
| 14:18:38 | efried | fwiw the date stamp implies that it is. | |
| 14:19:13 | efried | everything /usr/lib/libvirt* has a now-ish date stamp | |
| 14:19:35 | efried | and yes, unstacking removes everything /usr/lib/libvirt* | |
| 14:19:51 | sean-k-mooney | ok so if you run "pkg-config libvirt ; echo $?" | |
| 14:20:05 | sean-k-mooney | im guessin git now prints a non 0 value right | |
| 14:20:38 | sean-k-mooney | if so can you cd in /opt/stack/libvit and do sudo make install | |
| 14:20:42 | sean-k-mooney | then run it again | |
| 14:21:38 | efried | yes, confirm all of that. | |
| 14:22:10 | sean-k-mooney | so if you run it after you do make install is it 0 or non 0 | |
| 14:22:18 | efried | 0 | |
| 14:22:34 | sean-k-mooney | ok and if you manually to the pip install of libvirt-python? | |
| 14:22:35 | efried | and the checksum of the lib is the same as on the other system where I didn't unstack and manually `make install` | |
| 14:22:48 | sean-k-mooney | ya it should be | |
| 14:23:06 | efried | same error: | |
| 14:23:06 | efried | /usr/bin/pkg-config --print-errors --atleast-version=0.9.11 libvirt | |