| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-07-20 | |||
| 17:00:41 | bbansal | https://review.openstack.org/#/c/485406 | |
| 17:01:12 | mriedem | bbansal: there is a bot that syncs the global-requirements changes to nova | |
| 17:01:24 | mriedem | like this https://review.openstack.org/#/c/455970/ | |
| 17:01:49 | ildikov | mriedem: checking | |
| 17:01:51 | bbansal | okay, so i need to wait until changes are updated on all projects | |
| 17:03:43 | bbansal | @mriedem: yesterday u filed a bug related to n-net if u remember, can u please provide change set again | |
| 17:04:42 | bbansal | or the bug id | |
| 17:05:29 | sdague | cdent: yeh, I did, I just wanted to get things passing first. | |
| 17:05:44 | sdague | I'll fix the commit message before moving it out of WIP | |
| 17:08:41 | larsks | dansmith: btw, i just rebased https://review.openstack.org/#/c/467699/, which we had previously discussed. I see a jenkins failure but pretty sure it's unrelated. | |
| 17:09:32 | dansmith | larsks: okay, I'm trying to think of who the best person to review that is | |
| 17:09:34 | dansmith | mriedem: ^ | |
| 17:18:35 | ildikov | mriedem: if you got a moment, I got lost a bit on the VolumeUnattached stuff in remove check_detach | |
| 17:22:57 | ildikov | mriedem: in check_detach the VolumeUnattached exception should not be raised without multi-attach and I think even with that it would mean that something strange is going on | |
| 17:23:19 | ildikov | mriedem: without it we hit InvalidVolume | |
| 17:26:05 | openstackgerrit | Kashyap Chamarthy proposed openstack/nova master: libvirt/driver.py: Set cache value for Cinder volume post-migration https://review.openstack.org/485752 | |
| 17:26:20 | mriedem | bbansal: https://review.openstack.org/#/c/485300/ | |
| 17:27:31 | mriedem | dansmith: larsks: does this go into config drive / metadata service API? | |
| 17:27:37 | dansmith | yeah | |
| 17:27:48 | bbansal | thanks | |
| 17:28:08 | mriedem | hmm | |
| 17:28:20 | mriedem | so just our usual old random shove whatever we want in there i guess? | |
| 17:28:31 | larsks | mriedem: note that this change preserve existing behavior. | |
| 17:28:48 | dansmith | mriedem: yeah, I think it's pretty compatible | |
| 17:28:48 | larsks | ...and implements more correct behavior for those who want it. | |
| 17:29:26 | dansmith | mriedem: we've been doing no more fine-grained changes than one per release right? in terms of versions, I mean | |
| 17:29:33 | dansmith | and we've not really been super consistent on that, AfAIK | |
| 17:30:37 | mriedem | we haven't been versioning the metadata api at all when network_data.json changes as far as i know | |
| 17:30:44 | mriedem | there was some other change or two earlier in pike that didn't have a version bump | |
| 17:30:46 | dansmith | oh okay | |
| 17:30:55 | mriedem | filed under bug fixes | |
| 17:30:59 | dansmith | even better then | |
| 17:31:00 | dansmith | "better" | |
| 17:31:14 | mriedem | right. if this were the rest api, it would be a response body change | |
| 17:31:22 | mriedem | *compute api | |
| 17:34:56 | mriedem | left comments | |
| 17:35:00 | mriedem | this isn't so much a bug as a wishlist item | |
| 17:35:04 | mriedem | but i didn't -1 | |
| 17:36:30 | dansmith | mriedem: did you read the bug? | |
| 17:37:10 | dansmith | IIRC, you can't properly configure redhat-style interface definitions with just the global list | |
| 17:37:39 | dansmith | if you use the smashed together version, then you break resolution when you bring interfaces down | |
| 17:37:43 | dansmith | *some interfaces I mean | |
| 17:38:33 | mriedem | ok, i didn't -1 | |
| 17:38:33 | larsks | mriedem: I claim this is absolutely a bug. | |
| 17:38:43 | mriedem | i left some concerns | |
| 17:38:48 | dansmith | mriedem: it's totally not wishlist, IMHO :) | |
| 17:38:52 | mriedem | i have to run for a bit | |
| 17:39:00 | mriedem | ok change the severity in LP | |
| 17:39:01 | mriedem | gotta run | |
| 17:55:28 | trinaths1 | Hi, I have installed libvirt 1.3.5 and Qemu 2.6 in compute. When I start libvirtd service, I get this error http://paste.openstack.org/show/616064/ please help me solve this issue. | |
| 17:59:34 | sean-k-mooney | trinaths1: did you also install the correct version of libvirt python to match your libvirt. it looks like there is an issue with your libvirt config/path also as it is trying use qemu 2.4 not 2.6 | |
| 18:00:24 | jaypipes | dansmith: what the... ? http://logs.openstack.org/42/446242/41/check/gate-nova-tox-functional-ubuntu-xenial/30ad602/testr_results.html.gz | |
| 18:00:45 | jaypipes | dansmith: I swear, gate functional tests are getting randomer and randomer failures :( | |
| 18:00:49 | dansmith | jaypipes: access was denied to this resource | |
| 18:00:50 | dansmith | duh | |
| 18:00:56 | jaypipes | heh | |
| 18:01:06 | trinaths1 | sean-k-mooney: ok. is there any misconfiguration i'm doing ? | |
| 18:01:15 | cdent | that’s the thing where requests to placement go to compute (for reasons unknown) | |
| 18:01:55 | sean-k-mooney | trinaths1: did you build qemu and libvirt yourself. if you are using packages it shoudl have fix this for you | |
| 18:01:57 | cdent | fixtures not playing well with one another | |
| 18:02:52 | trinaths1 | sean-k-mooney: yes. I have build from git source code. qemu 2.4 and Libvirt 1.3.5 | |
| 18:04:44 | trinaths1 | sean-k-mooney: but i have installed libvirt-python using dnf. on fedora 23 | |
| 18:05:39 | sean-k-mooney | trinaths1: does virsh capabilities work for you? | |
| 18:07:29 | sean-k-mooney | trinaths1: actully was that form the libvirtd unit in journalctl? http://paste.openstack.org/show/616064/ | |
| 18:08:06 | sean-k-mooney | trinaths1: i taught the log was from nova at first hench question regarding libvirt python binding | |
| 18:08:24 | trinaths1 | sean-k-mooney: no. it did give complete details. when I start nova-compute, I get this error: http://paste.openstack.org/show/616067/ | |
| 18:08:48 | sean-k-mooney | trinaths1: this actully looks like the libvirt version you are using is not compatible with the qemu you have installed. | |
| 18:09:36 | openstackgerrit | Ildiko Vancsa proposed openstack/nova master: Implement new attach Cinder flow https://review.openstack.org/330285 | |
| 18:09:36 | openstackgerrit | Ildiko Vancsa proposed openstack/nova master: Remove check_detach https://review.openstack.org/446671 | |
| 18:10:10 | trinaths1 | sean-k-mooney: oops! please guide me to understand proper qemu for libvirt 1.3.5 | |
| 18:10:44 | sean-k-mooney | trinaths1: well it looks like you have 2 issues 1st No compute node record for host fedora23-1046: ComputeHostNotFound_Remote: Compute host fedora23-1046 could not be found. | |
| 18:11:18 | trinaths1 | sean-k-mooney: i resolved it now. its working fine now. | |
| 18:11:36 | trinaths1 | sean-k-mooney: but 2nd issue is still pending .. | |
| 18:11:37 | sean-k-mooney | trinaths1: was that caused by cells? | |
| 18:11:50 | trinaths1 | sean-k-mooney: no. we are using mitaka compute. | |
| 18:17:34 | trinaths1 | sean-k-mooney: see this, https://wiki.openstack.org/wiki/LibvirtDistroSupportMatrix I think 1.3.5 can be compatible with qemu 2.4 | |
| 18:18:19 | trinaths1 | sean-k-mooney: I aslo tried libvirt 1.3.5 with qemu 2.6 . it doesn't work well | |
| 18:19:13 | sean-k-mooney | trinaths1: it may be compatible on x86 not sure about arm. it may be down to how you compiled qemu | |
| 18:19:49 | trinaths1 | sean-k-mooney: ok | |
| 18:20:12 | sean-k-mooney | are you deploying with devstack out of interest? | |
| 18:20:42 | trinaths1 | sean-k-mooney: I'm using manual installation. | |
| 18:20:56 | trinaths1 | sean-k-mooney: not using devstack | |
| 18:21:34 | sean-k-mooney | ah we wrote a devstack plugin to compile libvirt and qemu from source but i have not tested it on arm or we qemu and libvirt versions that old. https://github.com/openstack/devstack-plugin-libvirt-qemu. | |
| 18:22:19 | sean-k-mooney | trinaths1: it might still be good to compare how we do the compilation https://github.com/openstack/devstack-plugin-libvirt-qemu/blob/master/devstack/libs/libvirt and see if you did anything different | |
| 18:22:49 | trinaths1 | sean-k-mooney: oh ok. but as you have said above. i require libvirt 1.3.5 and qemu 2.4 | |
| 18:23:52 | openstackgerrit | Merged openstack/nova master: Make key_pairs use check_deltas() for quota https://review.openstack.org/477699 | |
| 18:24:14 | sean-k-mooney | trinaths1: do you need exactly those version or are they the minium. fedroas rpms should actully be newer then that but if you are trying to reporduce an issue that wont help you. | |
| 18:25:31 | trinaths1 | sean-k-mooney: agree. but libvirt 1.3.5 is mandatory. | |
| 18:25:48 | trinaths1 | sean-k-mooney: but qemu can be something different above 2.2 | |
| 18:27:53 | sean-k-mooney | trinaths1: it might be worth installing qemu from dnf then instead and see if it resolves the issue. | |
| 18:27:58 | sean-k-mooney | actully | |
| 18:28:14 | sean-k-mooney | trinaths1: do you have selinux enabled | |
| 18:28:30 | sean-k-mooney | its a long shot but that could be blocking the request | |
| 18:29:29 | sean-k-mooney | trinaths1: your main error though is internal error: No JSON parser implementation is available | |
| 18:29:43 | trinaths1 | sean-k-mooney: SELinux status: disabled | |
| 18:29:48 | sean-k-mooney | trinaths1: that is why i taught it might be related to how you compiled libvirt | |
| 18:30:11 | trinaths1 | sean-k-mooney: yes. True said. but I find no help from google on this. | |
| 18:30:26 | trinaths1 | sean-k-mooney: so finally i have come here. | |
| 18:30:33 | sean-k-mooney | trinaths1: if you still have it i would check the configure output ans see if there is is any reference to json | |
| 18:32:49 | trinaths1 | sean-k-mooney: sure. I'm currently rebuilding libvirt fresh with qemu 2.4 refering the devstack plugin link you have provided. | |