| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-07-23 | |||
| 13:35:46 | dansmith | mriedem: I guess I would think returning shell results for missing services would be super low priority to me | |
| 13:36:09 | dansmith | mriedem: if a cell is down and we just return the services we can find, I would think the admins would be fine with that given they know the cell is down | |
| 13:36:11 | mriedem | dansmith: as in might as well make them a single microversion | |
| 13:36:33 | mriedem | them = nova list & service list | |
| 13:36:47 | dansmith | mriedem: we are already returning partial results for service list if a cell is down right? | |
| 13:36:54 | mriedem | yes | |
| 13:37:06 | mriedem | tssurya did the fix for that as a bug | |
| 13:37:16 | tssurya | not partial results, | |
| 13:37:20 | tssurya | like skips that cell | |
| 13:37:21 | dansmith | I'm saying, focus on instance list returning shell instances to patch up the instance list, but not care about service list | |
| 13:37:28 | dansmith | tssurya: that's what I mean by partial results | |
| 13:37:34 | dansmith | just omitting the downed cell | |
| 13:37:37 | tssurya | ah yes sorry, | |
| 13:37:57 | dansmith | I'm not sure why we really need to try to patch up the service list tbh | |
| 13:38:04 | dansmith | and I imagine it's going to be hard | |
| 13:38:15 | mriedem | you can patch up the compute services, but that's about it | |
| 13:38:38 | tssurya | yea just the host and binary fields | |
| 13:38:43 | mriedem | i don't think it's all that hard. the value question is the question | |
| 13:38:59 | dansmith | I always get this confused, but aren't the host mappings done by compute node and not service? | |
| 13:39:18 | mriedem | should be the service | |
| 13:39:47 | dansmith | it's by compute node by default, | |
| 13:39:48 | tssurya | I guess its compute-node for normal cases and service for ironic ? | |
| 13:39:56 | dansmith | I added by-service recently | |
| 13:40:14 | mriedem | sure, but https://github.com/openstack/nova/commit/64e76de43dc55e584c100005fa60da50dd06d352 | |
| 13:40:15 | mriedem | oops | |
| 13:40:20 | mriedem | HostMapping.get_by_host(ctxt, compute.host) | |
| 13:40:33 | mriedem | the host mappings are still 1:1 with the nova-compute service host | |
| 13:40:35 | dansmith | ah, compute.host is service host yeah? | |
| 13:40:37 | dansmith | okay | |
| 13:40:41 | mriedem | there isn't a host mapping per compute.hypervisor_hostname | |
| 13:40:44 | mriedem | which is the 1:M thing | |
| 13:40:46 | dansmith | right right okay | |
| 13:41:13 | dansmith | I was thinking it would be hard if it was one mapping per node, but fair enough | |
| 13:41:23 | dansmith | anyway, it just seems like really low prio to me | |
| 13:41:27 | dansmith | brb | |
| 13:41:32 | mriedem | given ^ why did we need the discover by service again? | |
| 13:42:10 | mriedem | oh b/c ironic doesn't have the nodes yet | |
| 13:42:15 | mriedem | by the time we need to discover the hosts | |
| 13:42:58 | mriedem | anyway, i agree doing shell results for service list is also low(er) priority so that's why i figured it would be part of the single microversion for doing nova list/show shell results | |
| 13:43:22 | tssurya | okay, so we do all of that in the same microversion then | |
| 13:43:45 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Send soft_delete from context manager https://review.openstack.org/476459 | |
| 13:44:35 | tssurya | and for now, we just decide to go with dansmith's migration ? thb, I don't have any particular preference between the two patches | |
| 13:44:47 | tssurya | so both works for us | |
| 13:45:50 | dansmith | mriedem: well my point about being lower priority meant I wouldn't want to hold up the instance side for the service side | |
| 13:46:36 | mriedem | dansmith: sure, she has the service one done* | |
| 13:46:53 | mriedem | and i'm saying i figured just leave it until the instance one is done and make them a single microversion | |
| 13:47:03 | mriedem | then that is the "down cell" microversion | |
| 13:47:10 | mriedem | but that's just how i've always been thinking this would happen | |
| 13:47:27 | dansmith | oh I see, | |
| 13:47:33 | dansmith | I had assumed the opposite | |
| 13:47:44 | dansmith | (that the instance one was done already) | |
| 13:47:56 | dansmith | my fault for not being up to date | |
| 13:48:14 | tssurya | nova list is held up for now because it would be neater if this goes in: https://review.openstack.org/#/q/topic:bp/api-extensions-merge-rocky+(status:open+OR+status:merged) | |
| 13:48:28 | mriedem | and we'll be lucky to just get ^ in to rocky | |
| 13:48:36 | mriedem | given our current merge rates | |
| 13:48:38 | dansmith | oye | |
| 13:48:53 | mriedem | it's taken a week to land code right now | |
| 13:49:07 | dansmith | yarp | |
| 13:49:16 | mriedem | so i'm really only mostly interested in pushing things across the line this week that is already ready to go and has had a lot of review already | |
| 13:49:18 | mriedem | personally, | |
| 13:49:33 | mriedem | plus i'm going to have a shitty week with some personal stuff going on so won't be fully available | |
| 13:50:15 | pooja_jadhav | dansmith: hello | |
| 13:50:26 | dansmith | mriedem: ack | |
| 13:50:29 | dansmith | pooja_jadhav: hi | |
| 13:50:41 | efried | Scheduler meeting in ten minutes in #openstack-meeting-alt | |
| 13:50:52 | tssurya | mriedem: ack thanks a lot for all the time an review doen on this spec (if I haven't yet thanked you) | |
| 13:51:21 | tssurya | dansmith: thanks to you too for all the help and reviews | |
| 13:51:43 | pooja_jadhav | dansmith: Actually, gone through your patch, https://review.openstack.org/#/c/580720/. Now testing the move operations (resize, unshelve, evacuate and migrate). | |
| 13:52:54 | openstackgerrit | Dan Smith proposed openstack/nova master: Online data migration for queued_for_delete flag https://review.openstack.org/584504 | |
| 13:53:21 | dansmith | tssurya: mriedem: fixed py35 issue in the test ^ | |
| 13:53:23 | mriedem | pooja_jadhav: are there problems or you're just letting us know you're testing? | |
| 13:53:32 | mriedem | pooja_jadhav: also, thanks for testing this stuff | |
| 13:53:34 | tssurya | dansmith: ack | |
| 13:53:53 | pooja_jadhav | dansmith, mriedem: I have 2 nodes, I am facing issue for migrate API. Can u tell me what are the pre requisites? | |
| 13:54:08 | mriedem | pooja_jadhav: cold migrate? | |
| 13:54:45 | mriedem | unless the nodes are on the same shared storage pool you have to configure ssh access for the nova user between the nodes | |
| 13:55:02 | pooja_jadhav | mriedem, dansmith: for unshelve, there is issue (DISK_GB record is added after unshelve the instance). | |
| 13:55:13 | mriedem | pooja_jadhav: added where? | |
| 13:55:22 | mriedem | to placement? | |
| 13:55:22 | dansmith | mriedem: you have a patch for that right? | |
| 13:55:25 | mriedem | dansmith: yeah | |
| 13:55:45 | pooja_jadhav | mriedem: yes, in the allocation table | |
| 13:56:02 | pooja_jadhav | can I provide the details, I have tested? | |
| 13:56:30 | mriedem | unshelve should be handled here https://review.openstack.org/#/c/583715/3/nova/conductor/manager.py@806 | |
| 13:57:16 | mriedem | i could update the functional test in that change to run a shelve/unshelve routine to make sure DISK_GB still doesn't show up in placement allocations | |
| 13:58:07 | pooja_jadhav | mriedem: cold/live migrate (both trying to test) | |
| 13:58:27 | mriedem | damn i have to rebase this series because https://review.openstack.org/#/c/583646/ was rebased | |
| 13:58:35 | mriedem | pooja_jadhav: https://docs.openstack.org/devstack/latest/guides/multinode-lab.html | |
| 13:58:38 | mriedem | if you're using devstack | |
| 13:59:52 | pooja_jadhav | mriedem: thanks, will check and let u know if the queries :) | |
| 14:00:04 | pooja_jadhav | yes, I am using devstack | |
| 14:00:26 | mriedem | that doc might be old | |
| 14:04:59 | pooja_jadhav | mriedem: Also tested resize and evacuate, but no allocation records get added for DISK_GB. | |
| 14:06:03 | pooja_jadhav | mriedem: But for evacuate, hypervisor-stats showing incorrect incorrect usages. | |
| 14:06:15 | openstackgerrit | Boxiang Zhu proposed openstack/nova stable/pike: Fix "instance snap min disk size err after resize instance" https://review.openstack.org/584770 | |
| 14:08:52 | mriedem | pooja_jadhav: we already know that hypervisor-stats has latent bugs for reporting disk based on the flavor and not if the server is volume-backed | |
| 14:09:08 | mriedem | that's what https://review.openstack.org/#/c/583646/ shows | |
| 14:09:15 | pooja_jadhav | mriedem: Ohk | |
| 14:10:26 | pooja_jadhav | mriedem: Have tested live/cold migrate then? for allocation of DISK_GB or not? | |
| 14:11:13 | mriedem | pooja_jadhav: https://review.openstack.org/#/c/583715/ functionally tests that for cold migration | |