| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-03-15 | |||
| 13:03:12 | lpetrut | mdbooth: do you know by any chance if there are any other iSCSI drivers that would return different connection info among hosts (i.e. have a different conn info on the destination)? | |
| 13:03:37 | mdbooth | lpetrut: Potentially a lot, I think | |
| 13:03:47 | mdbooth | That particular issue has been around for ages | |
| 13:03:54 | lpetrut | for FC it would look different but that's fine as disconnect_volume is a noop for FC in our case | |
| 13:04:15 | mdbooth | The new issue is that there's one driver where initialize_connection doesn't return the same value if you call it a second time for the *same* host | |
| 13:04:37 | lpetrut | got it | |
| 13:05:50 | mdbooth | lpetrut: This was the original bug: https://bugs.launchpad.net/nova/+bug/1475411 | |
| 13:05:52 | openstack | Launchpad bug 1475411 in nova (Ubuntu Trusty) "During post_live_migration the nova libvirt driver assumes that the destination connection info is the same as the source, which is not always true" [Undecided,Fix released] | |
| 13:06:05 | mdbooth | That's obviously from 2015 | |
| 13:06:54 | mdbooth | Affects 3par, apparently: https://bugs.launchpad.net/nova/+bug/1288039 | |
| 13:06:55 | openstack | Launchpad bug 1288039 in OpenStack Compute (nova) "live-migration cinder boot volume target_lun id incorrect" [High,Fix released] - Assigned to Anthony Lee (anthony-mic-lee) | |
| 13:08:14 | mdbooth | Anyway, if Hyper-V uses that cinder driver, my expectation is that my new change will fix it | |
| 13:08:27 | lpetrut | thanks for the link, indeed, that should fix it | |
| 13:09:36 | lpetrut | I was thinking about submitting a bug report against our driver for posterity reasons, marking your patch as a fix | |
| 13:10:54 | lpetrut | I was not aware of the 3par issue, high chance is that we were leaking volume connections and nobody noticed | |
| 13:11:56 | mdbooth | Is anybody aware of a way to make devstack do git checkouts, install pip dependencies, and then stop? | |
| 13:23:37 | kashyap | mtreinish might know ^ | |
| 13:24:58 | cdent | mdbooth: is this in a session you're running by hand or something you want to automate | |
| 13:28:44 | cdent | mdbooth: if the former, probably just hack an exit in stack.sh | |
| 13:29:02 | ameeda | any help? Hello, I am trying to upload image from server1 to openstack "server2" using rest api. when I did that I got error message "504 Gateway Time-out The server didn't respond in time." | |
| 13:29:22 | ameeda | I set glance conn timeout to 0 , with no lucks | |
| 13:30:04 | cdent | mdbooth: also grep on TRACK_DEPENDS; let's you bail out after install | |
| 13:30:32 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Cleanup tempest-dsvm-cells-rc blacklist https://review.openstack.org/553401 | |
| 13:30:32 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Make nova-cells-v1 run with neutron https://review.openstack.org/549789 | |
| 13:31:16 | tssurya | mriedem : owe you one for the NewtonCellCheck patch, thanks | |
| 13:31:43 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Cleanup tempest-dsvm-cells-rc blacklist https://review.openstack.org/553401 | |
| 13:31:49 | mriedem | tssurya: np | |
| 13:32:21 | ameeda | mriedem: Hello :) , could you please help me with my issue ? | |
| 13:32:59 | mriedem | ameeda: um, check the logs | |
| 13:33:36 | ameeda | mriedem: let me check , thanks | |
| 13:36:28 | mdbooth | cdent: Automate. I'm trying to build a base devstack image with all the config-agnostic stuff pre-done. Will look at TRACK_DEPENDS, thanks. | |
| 13:36:59 | kholkina | hi all! could you please review the spec? https://review.openstack.org/#/c/547964 | |
| 13:40:20 | claudiub|2 | mdbooth: sorry, I was out for a bit. Will look at it. | |
| 13:41:05 | mdbooth | claudiub|2: Thanks. lpetrut already looked, btw. | |
| 13:44:30 | ameeda | mriedem: I check this log /var/log/glance/api.log, I didn't find anything helpful , also this command journalctl -u openstack-glance-api.service -f | |
| 13:45:51 | SamYaple | ameeda: how did you deploy glance? uwsgi? apache? eventlet? | |
| 13:46:53 | mriedem | kholkina: -2 on that idea, you can already update user-data via rebuild now | |
| 13:47:03 | mriedem | a lot of thought and discussion already went into adding that to rebuild | |
| 13:47:05 | mriedem | see http://lists.openstack.org/pipermail/openstack-operators/2017-October/014309.html | |
| 13:47:12 | ameeda | SamYaple: I create python project to use contact with openstack via rest api, I use http requests as rest client | |
| 13:47:47 | mriedem | ameeda: this isn't really the channel for this type of question - try #openstack or #openstack-glance | |
| 13:49:14 | ameeda | everything working on my project, also I test the project on packstack, its working, but this case have to copy image file from server1 to openstack server, so the image file not exists on the same openstack | |
| 13:50:41 | ameeda | mriedem: I will try the other channels, thank you | |
| 13:52:15 | efried | jaypipes, cdent: Opinion on https://review.openstack.org/#/c/548975/5/nova/scheduler/client/report.py@592 ? | |
| 13:52:38 | cdent | efried: explicit is better than implicit | |
| 13:52:52 | cdent | some version ought to be listed there | |
| 13:52:58 | kholkina | mriedem, the spec means to update user-data on booted instance, without rebuild. It may be useful to provide a cloud-config | |
| 13:54:15 | efried | cdent: Even if it's '1.0'? | |
| 13:54:18 | jaypipes | efried: yeah, agree with alex_xu_ it would be good to be consistent there. | |
| 13:54:25 | cdent | efried: yes | |
| 13:54:30 | jaypipes | efried: no biggie though, really | |
| 13:55:01 | efried | cdent: Okay. Note that there's places we don't do that, elsewhere in the report client. A big enough deal to put up a patch to fill those in? | |
| 13:55:08 | cdent | no | |
| 13:55:33 | cdent | It's, as jay says, no biggie, but if you're adding something new may as well make the version explicit, even if it is the base | |
| 13:55:59 | alex_xu_ | that method is only used for create root rp? | |
| 13:56:02 | edleafe | It also looks consistent in that block to have a version in both | |
| 13:56:30 | cdent | that will help make things more clear if at some point you do need to add a parent or something there and it doesn't work and you're all wtf, why is this not taking my parent, oh the versin | |
| 13:57:10 | efried | fixing. | |
| 13:57:13 | efried | alex_xu_: yes | |
| 13:57:26 | efried | oh | |
| 13:57:29 | efried | maybe not. Stand by. | |
| 13:58:05 | efried | Hah, you're right, it can assign a parent. So my change was blatantly wrong on that point. Good catch, alex_xu_. Fixing... | |
| 13:58:59 | openstackgerrit | Eric Fried proposed openstack/nova master: Stop assuming initial provider generation is 0 https://review.openstack.org/548975 | |
| 13:59:01 | efried | alex_xu_, jaypipes, cdent: Done ^ | |
| 13:59:50 | ratailor | alex_xu_, jaypipes could you pl have a look at https://review.openstack.org/#/c/401009/ | |
| 14:03:07 | belmoreira | dansmith, mriedem in my cellsV1/V2 migration plan I have some issues that I would like to discuss you. Let me know when you are available | |
| 14:08:23 | openstackgerrit | Rajesh Tailor proposed openstack/nova master: Reset instance status on invalid host https://review.openstack.org/401009 | |
| 14:08:27 | dansmith | belmoreira: go for it | |
| 14:08:57 | belmoreira | thanks dansmith | |
| 14:09:03 | belmoreira | We use the same cloud to deploy VMs and baremetal. Baremetal is dedicated to few cells. | |
| 14:09:22 | belmoreira | When consolidating the scheduler at top level (not per cell as in cellsV1) not seeing how can dedicate schedulers to ironic cells because they need the baremetal_filters enabled. | |
| 14:09:45 | dansmith | belmoreira: the baremetal filters are deprecated | |
| 14:10:27 | dansmith | belmoreira: the way of the light is custom resource classes for the ironic nodes, and flavors configured to give you the right thing when you ask, which makes scheduling uniform | |
| 14:11:04 | belmoreira | I missed that | |
| 14:11:07 | mriedem | there was a caveat with dropping the ironic host manager for the scheduler, would have to dig it back up, | |
| 14:11:32 | mriedem | but i think it's that not all filter scheduler filters will be able to handle baremetal instances, numatopologyfilter specifically i think | |
| 14:12:42 | belmoreira | I'm finding the release notes about this. thanks | |
| 14:12:54 | mriedem | i'll be damned if i can find the way to those docs in ironic via their homepage | |
| 14:13:15 | belmoreira | :) | |
| 14:13:18 | mriedem | belmoreira: https://docs.openstack.org/ironic/latest/install/configure-nova-flavors.html#scheduling-based-on-resource-classes | |
| 14:13:21 | mriedem | search box ftw | |
| 14:13:44 | alex_xu_ | mriedem: does make sense to move the get_compute_nodes before the state change https://review.openstack.org/#/c/401009/9/nova/compute/api.py@4222 ? | |
| 14:13:57 | alex_xu_ | it pretty looks like an API layer check. | |
| 14:14:05 | belmoreira | thanks | |
| 14:14:43 | mriedem | alex_xu_: i wouldn't be opposed to that, replied | |
| 14:14:55 | alex_xu_ | mriedem: ok | |
| 14:15:18 | mriedem | belmoreira: https://github.com/openstack/nova/commit/c99fc64271f41a028c75e96348813f673d04e245 | |
| 14:16:26 | mriedem | NUMATopologyFilter isn't enabled by default, but i know a lot of people use it | |
| 14:17:19 | mriedem | if we had someone that was familiar enough with both nova and ironic (cough jroll johnthetubaguy mgoddard_ cough) we could have them play around with updating an ironic CI job to not use the ironic host manager and baremetal filters and see what breaks | |
| 14:18:20 | mgoddard_ | mriedem can I offer you a throat sweet? | |
| 14:18:31 | mriedem | wtf | |
| 14:18:39 | mgoddard_ | for your cough | |
| 14:18:40 | mriedem | is that british for "i'm going to slit your throat?" | |
| 14:18:42 | mriedem | oh | |
| 14:19:08 | mgoddard_ | oh wow, crossed wires there | |
| 14:19:26 | mriedem | i consider phlegm a feature so no thanks | |
| 14:19:41 | dansmith | PaaS - Phlegm as a Service? | |
| 14:19:42 | mgoddard_ | I could take a look at this | |
| 14:20:12 | mriedem | i'm having trouble remembering off the top of my head what it was about the NUMATopologyFilter that makes it not work with ironic | |
| 14:20:23 | dansmith | mriedem: that needs fixing if so | |