| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-04-11 | |||
| 10:28:09 | BobBall | johnthetubaguy / artom: Found the problem - it's relating to Ia0869dc6f7f5bd347ccbd0930d1d668d37695a22 (device tag disappearing) - because nova re-added the device tag in the 2.42 version but novaclient doesn't support that version, we have no way of using device tagging currently. | |
| 10:28:15 | jsheeren | anyone know if it is possible to limit image A to be only able to use flavor X and Y, image B only flavor Z? | |
| 10:28:37 | johnthetubaguy | BobBall: oh I forgot about that being for tags | |
| 10:29:02 | BobBall | Not sure what the fix is currently -I did try hacking novaclient but it still can't cope with device tags for block devices... | |
| 10:29:24 | johnthetubaguy | jsheeren: there are various image properties that can help for specific requirements, not sure there is a blanket only flavor X thing. | |
| 10:29:50 | johnthetubaguy | BobBall: can't you raise the max supported API version up to 2.42? | |
| 10:32:09 | BobBall | works for nics but not for block devices (volume_id is not recognised) | |
| 10:35:46 | johnthetubaguy | BobBall: I am curious what the API tests are doing, I don't remember the context around that stuff | |
| 10:36:26 | BobBall | :P | |
| 11:24:00 | sfinucan | vladikr, sahid: Could you guys take a look at some point, please? https://review.openstack.org/#/q/topic:bump-min-libvirt+status:open | |
| 11:32:28 | jsheeren | johnthetubaguy: thanks | |
| 11:32:43 | openstackgerrit | Sergey Nikitin proposed openstack/nova master: Added PCI NUMA policies https://review.openstack.org/390520 | |
| 11:33:58 | vladikr | sfinucan, sure. will drop off my kids to school and then | |
| 11:43:17 | artom | BobBall, if you pin the version to 2.32 it should work for both nic and bdm tags | |
| 11:43:25 | artom | Until we merge the 2.42 novaclient change | |
| 11:48:28 | gaurangt | hi, I'm facing issue in the nova instance creation and it is failing with following error in nova-compute.log | |
| 11:48:30 | gaurangt | Failed to create resource provider record in placement API for UUID | |
| 11:48:39 | gaurangt | The requested URL /placement/resource_providers was not found on this server | |
| 11:49:11 | gaurangt | I actually added the <Directory> section in nova-placement-api.conf, but still the same issue. | |
| 11:50:03 | gaurangt | any idea what might be going wrong here? | |
| 11:50:44 | gaurangt | nova-placement-api.conf looks like this - http://paste.openstack.org/show/606143/ | |
| 11:51:41 | cdent | gaurangt: "/nova-placement-api" should be "/placement" in both places it shows up | |
| 11:51:56 | cdent | actually the Alias is wrong in other ways too, just a sec | |
| 11:53:14 | cdent | gaurangt: we're you following some docs to set this up? | |
| 11:53:58 | gaurangt | cdent, yes, I'm following the official OpenStack guide | |
| 11:53:59 | gaurangt | https://docs.openstack.org/ocata/install-guide-rdo/nova-controller-install.html | |
| 11:55:36 | cdent | gaurangt: it looks like the rdo packages may be configuring apache somewhat incorrectly | |
| 11:56:00 | gaurangt | cdent, oh | |
| 11:56:04 | cdent | changes the <Location ..> entry to be <Location /placement> and reload apache and see what happens | |
| 11:56:13 | gaurangt | cdent, what should be the workaround here? | |
| 11:56:21 | cdent | what I just said | |
| 11:56:38 | gaurangt | what about this - Alias /nova-placement-api /usr/bin/nova-placement-api? | |
| 11:56:58 | gaurangt | just above the <Location..> | |
| 11:57:46 | cdent | there too: Alias /placement /usr/bin/nova-placement-api | |
| 11:58:14 | gaurangt | cdent, hmm ok | |
| 11:59:06 | gaurangt | any nova service needs to be restarted? | |
| 11:59:13 | cdent | no just apache | |
| 11:59:30 | gaurangt | ok | |
| 11:59:52 | gaurangt | ah, now it says - You don't have permission to access /placement/resource_providers | |
| 12:02:01 | gaurangt | cdent, anything still wrong? | |
| 12:06:05 | cdent | move that <Directory> section outside of the VirtualHost | |
| 12:06:10 | cdent | gaurangt: ^ | |
| 12:06:50 | gaurangt | cdent, ok | |
| 12:10:26 | cdent | gaurangt: once you get this figured out, I'd suggest you submit a bug to the rdo packagers. It seems like there might some confusion between what's being set the apache configuration and what's being set in the service catalog. Either the packages or the docs are wrong | |
| 12:13:06 | gaurangt | cdent, thanks it worked | |
| 12:13:20 | cdent | gaurangt: ah, good, glad to hear it | |
| 12:13:23 | gaurangt | cdent, I will submit a bug now | |
| 12:13:30 | cdent | awesome! | |
| 12:13:57 | gaurangt | cdent, thanks for the help ! | |
| 12:50:06 | egonzalez | Hi, it's been a recent change in master who change the behaviour in nova service-list? in current master output is empty with no service registered | |
| 12:59:57 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: api-ref: Parameter verification for servers-actions (4/4) https://review.openstack.org/455573 | |
| 13:07:10 | jsheeren | hi again, i'm back with my question about pinning an image to a flavor | |
| 13:07:30 | jsheeren | we can set the min-disk and min-ram on the image, is there something similar with os_distro or os_type? | |
| 13:11:57 | BobBall | artom: Thanks but I don't think it will - nova client is rejecting the volume_id based on internal code, not checking the nova API version | |
| 13:13:59 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: api-ref: Example verification for servers-actions.inc https://review.openstack.org/454565 | |
| 13:15:01 | peter-hamilton | johnthetubaguy: do you have a minute to discuss cert validation? | |
| 13:23:20 | andymccr | cdent: are you about? | |
| 13:23:37 | cdent | andymccr: hi, yes | |
| 13:23:54 | andymccr | cdent: hey :) if you have a second - is there a reason why CONTENT_LENGTH has to be set to an int? https://github.com/openstack/nova/commit/6dd047a3307a1056077608fd5bc2d1c3b3285338 | |
| 13:24:16 | openstack | Launchpad bug 1681843 in OpenStack Compute (nova) "Nova-placement returns "ValueError: invalid literal for int() with base 10: ''"" [Undecided,New] | |
| 13:24:16 | andymccr | im having issues in that content_length is present, but empty so its failing with a valueError: https://bugs.launchpad.net/nova/+bug/1681843 | |
| 13:24:32 | andymccr | i figure the easiest fix is just to remove the int() (which works) but wanted to make sure im not overlooking anything | |
| 13:25:27 | cdent | andymccr: need to catch up and get my brain right, just a sec | |
| 13:25:41 | andymccr | cdent: no problem. lemme know if i can explain anything better/further | |
| 13:25:49 | peter-hamilton | jaypipes: do you have a sec to discuss cert validation? | |
| 13:26:13 | jaypipes | peter-hamilton: here on IRC? or in a hangout or something? | |
| 13:26:25 | peter-hamilton | jaypipes: here's fine, just some quick questions | |
| 13:26:35 | jaypipes | peter-hamilton: go for it. | |
| 13:27:02 | peter-hamilton | jaypipes: are you ok with adding two conf options, (1) to enable/disable cert validation, and (2) to provide a default list of trusted cert IDs | |
| 13:27:17 | peter-hamilton | jaypipes: the goal would be to smooth the migration path and user experience | |
| 13:27:25 | peter-hamilton | jaypipes: these were recommended by johnthetubaguy | |
| 13:27:39 | jaypipes | peter-hamilton: if that smooths the integration path, I'm all for it.' | |
| 13:28:32 | peter-hamilton | jaypipes: cool, i'm folding those in | |
| 13:28:37 | jaypipes | rock on brother. | |
| 13:30:53 | openstackgerrit | Andy McCrae proposed openstack/nova master: Don't force CONTENT_LENGTH to be an int() https://review.openstack.org/455710 | |
| 13:31:11 | cdent | andymccr: hmmm. that's icky behavior by nginx. what that check is trying to do is detect that there is a body with some length, without actually looking at the body, and then complain if there's no content-type header | |
| 13:31:40 | cdent | so if nginx is sending both, but with empty values, neither line is correct | |
| 13:32:33 | andymccr | cdent: agreed, although im confused as to why we need to force it to be an int? if there are no integer comparisons after that | |
| 13:33:11 | cdent | andymccr: probably because i was being overzealous: the spec says that if the header is present it's value is made up of DIGITS | |
| 13:33:20 | andymccr | ahh | |
| 13:33:54 | andymccr | cdent: any issue with removing the int? alternatively, we could get it, check it's not None and then make sure it's an int? | |
| 13:34:03 | andymccr | (or something slightly more efficient i assume) | |
| 13:34:10 | cdent | so yeah, your fix is probably okay, but if you're in there, addressing nginx'isms then you probably need to fix the content-type thing too | |
| 13:35:37 | cdent | a) drop the int b) if not environ.get('CONTENT_TYPE', ''): | |
| 13:35:55 | cdent | andymccr: that would get the correct behavior wouldn't it? | |
| 13:36:18 | andymccr | cdent: hmm thinking. | |
| 13:37:11 | andymccr | cdent: how it is now worked on content_type, because i guess it is defined, so they're both defined (but empty) - trying to think if there is another use case (that isnt this) where that would cause an issue | |
| 13:37:39 | mriedem | jaypipes: johnthetubaguy: sdague: at some point i need to talk with you about the os-interface response body and how we want that to look | |
| 13:37:49 | cdent | andymccr: the thing we are trying to prevent here is an empty content-type | |
| 13:38:05 | cdent | and I was thinking that the environ would only have the key if the header was present | |
| 13:38:09 | andymccr | cdent: so when CONTENT_LENGTH is not empty, make sure CONTENT_TYPE is also not empty | |
| 13:38:15 | cdent | yes | |
| 13:38:16 | andymccr | ohh i see what you mean | |
| 13:38:41 | andymccr | ok lemme fix! | |
| 13:40:15 | cdent | andymccr: thank you! | |
| 13:42:28 | frickler | did nova stop writing libvirt.xml into the instance-dir with ocata? or do I just need a specific option for it? | |
| 13:49:56 | openstackgerrit | Andy McCrae proposed openstack/nova master: Don't force CONTENT_LENGTH to be an int() https://review.openstack.org/455710 | |
| 13:58:19 | sdague | mriedem: I was looking through that this morning | |
| 13:58:40 | jaypipes | mriedem: sure, just gimme a few. need to restart my workstation.. | |
| 13:58:49 | sdague | mriedem: I guess the question is, what do we think people need to do with this? | |