| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-07-10 | |||
| 19:17:27 | dansmith | well, not the bit you said you didn't understand | |
| 19:17:30 | dansmith | does this help? https://review.openstack.org/#/c/471899/24/nova/api/openstack/compute/hypervisors.py | |
| 19:18:50 | mriedem | no | |
| 19:18:55 | mriedem | you want me to rename the parameter? | |
| 19:19:16 | mriedem | see https://review.openstack.org/#/c/471899/24/nova/api/openstack/compute/hypervisors.py@381 | |
| 19:23:31 | dansmith | I know, | |
| 19:23:58 | dansmith | I realize this is existing, I'm just saying it's super confusing/wrong to pass instances to a thing, calling them servers | |
| 19:24:42 | dansmith | just thought maybe we should clean that up since you're touching this code | |
| 19:24:44 | mriedem | so you want me to rename the formal parameter | |
| 19:24:53 | mriedem | i need you to tell me what you want me to do | |
| 19:24:54 | mriedem | :) | |
| 19:24:57 | mriedem | jaypipes: https://review.openstack.org/#/c/476632/ | |
| 19:25:42 | mriedem | i don't personally see how instances and servers is confusing, since instances feed the 'servers' resource in the API | |
| 19:25:47 | mriedem | are you thinking servers == hosts? | |
| 19:26:56 | dansmith | oh, I see, heh | |
| 19:26:58 | dansmith | yes | |
| 19:27:05 | mriedem | :/ | |
| 19:27:09 | dansmith | and also servers=services | |
| 19:27:14 | dansmith | more the latter I think | |
| 19:27:23 | dansmith | because I was already in the services frame of mind | |
| 19:27:30 | edleafe | ugh - I remember this naming confusion back from the early days with RAX | |
| 19:27:44 | mriedem | https://www.youtube.com/watch?v=D4O94hw3qt0 | |
| 19:27:44 | mriedem | oh yeah | |
| 19:27:47 | mriedem | that's the kind of mind i'm in | |
| 19:27:58 | edleafe | since they called their things "Cloud Servers", they insisted on calling an instance a "server" | |
| 19:28:11 | dansmith | servers is a stupid name for it, | |
| 19:28:26 | edleafe | dansmith: just one of the many arguments I lost in those days | |
| 19:28:42 | dansmith | but the ship has sailed, I just don't spend enough time in the api to remember all the time and the patch below is about _services_ | |
| 19:31:38 | mriedem | dansmith: if you're also good with the os-services change, can you +2 that so we can queue things up? | |
| 19:31:54 | mriedem | if i can get alex_xu to also be +2 on the series then we'll push it through | |
| 19:31:58 | dansmith | done | |
| 19:32:00 | mriedem | yay | |
| 19:36:04 | mriedem | now i can't get that song out of my head | |
| 19:37:56 | dansmith | mriedem: tbh I feel like I have a bit of review fatigue on this quota stuff | |
| 19:38:20 | dansmith | not sure I've seen an actual review from you on this one | |
| 19:38:29 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: claim resources in scheduler https://review.openstack.org/476632 | |
| 19:38:30 | mriedem | you haven't | |
| 19:38:33 | jaypipes | mriedem: fixed :( ^ | |
| 19:38:42 | dansmith | I guess we're not likely to see one from sdague in the near future either, eh? | |
| 19:39:12 | mriedem | probably not. i plan on reviewing that quotas one. | |
| 19:39:26 | jaypipes | mriedem: I will put a patch up renaming delete_allocation_for_instance() to delete_allocations_for_consumer() to remedy my need for grammatical correctness. | |
| 19:43:34 | sdague | dansmith: yeh, not this week | |
| 19:44:00 | sdague | I can get my head back into it next week, but I don't know which dates we're racing against now | |
| 19:44:19 | dansmith | yeah, need it before next week I think | |
| 19:52:26 | mriedem | sdague: feature freeze is july 27 | |
| 19:52:31 | mriedem | so we're less than 3 weeks from FF | |
| 19:53:09 | mriedem | jaypipes: a couple of comments in https://review.openstack.org/#/c/476632/ | |
| 19:53:25 | mriedem | jaypipes: dansmith: in a nutshell on that one, | |
| 19:53:47 | mriedem | i think it means we potentially have problems for bursts of instances in the same request with that change until we start doing retries in the scheduler | |
| 19:54:08 | mriedem | and/or we start randomizing which allocation request match we pick for a given compute node | |
| 19:55:39 | jaypipes | mriedem: do you want me to do retries in this patch? | |
| 19:55:49 | jaypipes | mriedem: I was working on that in a separate followup. | |
| 19:56:04 | jaypipes | mriedem: but if you're concerned about this, I suppose I could put it all in one patch. | |
| 19:56:11 | jaypipes | mriedem: was trying to keep the LOCC low. | |
| 19:56:16 | mriedem | a single change makes it harder to review for sure | |
| 19:56:17 | mriedem | yeah | |
| 19:56:42 | mriedem | i guess we assume no one is doing CD | |
| 19:57:12 | mriedem | jaypipes: how about line up the retries change on top, | |
| 19:57:13 | jaypipes | mriedem: answered you. | |
| 19:57:16 | mriedem | and then we'll get them in order | |
| 19:57:22 | jaypipes | mriedem: sure thing. | |
| 19:57:22 | mriedem | so there isn't like a week between changes | |
| 19:57:30 | jaypipes | mriedem: gimme about an hour. | |
| 20:00:31 | cdent | is this canon: “i guess we assume no one is doing CD” ? If so, life just got a lot easier. | |
| 20:00:49 | mriedem | no, it's not | |
| 20:01:19 | mriedem | dims is supposed to make huawei public cloud doing CD in the next couple of weeks :) | |
| 20:01:25 | cdent | do you take bribes? | |
| 20:02:54 | mriedem | https://docs.openstack.org/nova/latest/project_scope.html#upgrade-expectations | |
| 20:03:01 | mriedem | "In addition, we also support users upgrading from the master branch, technically, between any two commits within the same release cycle." | |
| 20:03:39 | cdent | oh yeah, I know, I was just thinking maybe the world had collapsed | |
| 20:30:03 | jaypipes | mriedem: so... problem. | |
| 20:30:17 | jaypipes | mriedem: I need to do the alternatives stuff before the retries. | |
| 20:30:29 | jaypipes | mriedem: because the retries need to loop over the alternatives. | |
| 20:30:34 | mriedem | right, b/c we need the alternatives to do retries | |
| 20:30:35 | mriedem | yup | |
| 20:30:49 | jaypipes | mriedem: I don't necessarily need to *return* (over RPC) the alternative yet though. | |
| 20:34:20 | mriedem | correctamundo | |
| 20:34:35 | mriedem | so it's like 3 changes on top of this, maybe 2 | |
| 20:34:40 | mriedem | 1. determine alternatives, | |
| 20:34:45 | mriedem | 2. retry on alternatives in scheduler | |
| 20:34:52 | mriedem | 3. pass alternatives to conductor | |
| 20:35:00 | mriedem | and then 4,5,6,7,8 do things with those in the cell | |
| 20:35:05 | jaypipes | yeah | |
| 20:38:50 | jaypipes | mriedem, dansmith: did we ever decide how many alternatives we wanted to return? | |
| 20:39:02 | jaypipes | mriedem, dansmith: limit to 10 per instance UUID maybe? | |
| 20:39:08 | dansmith | jaypipes: CONF.num_retries | |
| 20:39:13 | dansmith | times num_instances | |
| 20:39:13 | jaypipes | ah | |
| 20:39:14 | jaypipes | gotcha. | |
| 20:41:27 | edleafe | jaypipes: this was the consensus a couple of months ago: https://review.openstack.org/#/c/465767/ | |
| 20:41:51 | edleafe | default to CONF.num_retries | |
| 20:42:17 | edleafe | also, won't the order need to be different? | |
| 20:43:06 | edleafe | IOW, claim fails on "best" host, so you try claiming next one. Once you succeed, you select N alternates from the next N hosts in the weighed list that are in the same cell as the successful claimed host | |
| 20:43:18 | jaypipes | edleafe: the consensus was to create a new CONF option build_max_attempts? | |
| 20:44:42 | dansmith | that sounds like a terrible idea | |
| 20:44:45 | dansmith | num_retries and build_max_attempts is highly confusing | |
| 20:45:08 | dansmith | we have the former, I see no reason for a billionth conf var | |
| 20:45:54 | edleafe | jaypipes: sorry, it was supposed to be 'max_attempts' | |
| 20:47:47 | mriedem | plus host_subset_size | |
| 20:48:05 | openstackgerrit | Merged openstack/nova master: placement: scheduler uses allocation candidates https://review.openstack.org/476631 | |