| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-09-27 | |||
| 18:30:44 | mriedem | *migration | |
| 18:30:48 | mriedem | since we have to hit 2 different dbs | |
| 18:31:44 | imacdonn | personally I think the batch mechanism is a bit broken, at least as it's described in the comments | |
| 18:31:50 | mriedem | as i mentioned on your change, the 'Total Needed' column is misleading | |
| 18:31:50 | mriedem | http://logs.openstack.org/29/605329/2/check/neutron-grenade/2200365/logs/grenade.sh.txt.gz#_2018-09-27_11_17_32_536 | |
| 18:31:52 | imacdonn | but, as you said, we need Dan for that conversation | |
| 18:41:07 | mnaser | so has anyone ever thought what happens when we hit instance-ffffffff | |
| 18:43:10 | imacdonn | fsshhh ... that'll never happen | |
| 18:44:08 | melwitt | in case anyone is wondering about the failing ceph job, I'm trying out a fix here https://review.openstack.org/605833 | |
| 18:48:19 | mnaser | mordred: it's a bit of a difficult position but ideally figuring out what the best way to deal with cold/live migration and reworking it.. (openstack server migrate) | |
| 18:48:37 | mordred | yah. as you know, the sdk code for that is ... fun :) | |
| 18:48:47 | mnaser | mainly my issue was osc forces you to specify a host when its optional | |
| 18:48:47 | mordred | fwiw - mriedem is right - osc doesn't currently do version negotiation. once we start migrating it to sdk though, it'll pick up that ability | |
| 18:49:24 | mriedem | mnaser: easy: make --live just an option with no value, add --host (optional, takes a value), and add --cold | |
| 18:49:58 | mnaser | mriedem: but the not breaking scripts part i guess | |
| 18:50:02 | mriedem | or, let --live take a value for compat but proxy it to --host if --host isn't used | |
| 18:50:39 | mriedem | i'm not sure how you could specify --live w/o a host though if --live can take a host | |
| 18:50:41 | mriedem | gd CLIs | |
| 18:50:52 | mriedem | empty string? | |
| 18:50:58 | mriedem | pretty janky | |
| 18:51:25 | mriedem | openstack server migrate --live-but-with-no-host-seriously my_server | |
| 18:51:32 | mordred | mriedem: ++ | |
| 18:51:35 | mordred | that's totally the right answer | |
| 18:51:50 | mriedem | do i win something? | |
| 18:52:00 | mordred | you win this bucket of parts I just found | |
| 18:54:33 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add more documentation for online_data_migrations CLI https://review.openstack.org/605836 | |
| 18:54:42 | mriedem | imacdonn: efried: ^ does this make life better? | |
| 18:55:54 | imacdonn | only slightly, IMO | |
| 18:56:27 | mriedem | :( | |
| 18:57:12 | imacdonn | if those two rows don't need migration, then they couldn't be included in something named "Total Needed"......... ? | |
| 18:57:20 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/queens: nova-manage - fix online_data_migrations counts https://review.openstack.org/605839 | |
| 18:57:45 | mriedem | like i said, total needed is a bad title, | |
| 18:57:55 | imacdonn | if "Total Needed" means "Total Rows that exist that may or may not need it", we should see a lot less zeroes | |
| 18:57:56 | mriedem | i'm not sure if renaming that to Total Found breaks any kind of compat, | |
| 18:58:07 | mriedem | Total Candidates | |
| 18:58:09 | mriedem | something like that | |
| 18:58:28 | mriedem | no one should be parsing the output of this command for column headers and such anyway | |
| 18:58:46 | efried | hah | |
| 18:58:48 | imacdonn | I suspect that the existing migration methods may already interpret it inconsistently, but I haven't analysed it to confirm | |
| 18:59:42 | efried | yeah, to me, seeing needed=2/completed=0 feels like it should be an error | |
| 19:00:01 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/pike: nova-manage - fix online_data_migrations counts https://review.openstack.org/605840 | |
| 19:01:17 | efried | I mean, this is better than it was before, because there's at least *some* attempt to explain wtf is going on. | |
| 19:01:41 | efried | Is this dansmith's bailiwick btw? Something he would want to review? | |
| 19:05:45 | mriedem | i assume he would yes | |
| 19:06:22 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/ocata: nova-manage - fix online_data_migrations counts https://review.openstack.org/605842 | |
| 19:11:13 | mordred | mriedem, cfriesen: I just left a suggestion on the osc live migration patch about a way to make --live take an optional argument | |
| 19:12:54 | cfriesen | mordred: sweet, I think that's probably the best way to handle backwards compatibility | |
| 19:13:28 | mriedem | there are two, but i found it | |
| 19:14:25 | mriedem | i like that idea yeah | |
| 19:14:34 | mriedem | note you can't also cold migrate and specify a target host now... | |
| 19:14:53 | mriedem | so i'm not sure how that would play with this too | |
| 19:15:00 | mriedem | *can also | |
| 19:15:14 | imacdonn | I always thought that was weird .. that you can't do that (but can for live) | |
| 19:15:44 | imacdonn | (always => since icehouse days, at least) | |
| 19:16:06 | cfriesen | imacdonn: artificial OSC limitation | |
| 19:16:34 | cfriesen | imacdonn: the compute API lets you specify a host since 2.56 | |
| 19:16:34 | imacdonn | hmm | |
| 19:17:20 | cfriesen | (which is admittedly fairly new) | |
| 19:17:37 | imacdonn | right .. I was just checking ;) | |
| 19:18:16 | imacdonn | so "the struggle was real" when I last looked | |
| 19:18:28 | artom | I'm guessing splitting live and cold migration into different subcommands is no longer an option at this point, right? | |
| 19:18:37 | artom | Since they're, you know, fundamentally different operations? | |
| 19:19:20 | cfriesen | artom: to the end user, they're very similar. | |
| 19:19:36 | artom | cfriesen, you mean besides the fact that your workload goes down? | |
| 19:19:37 | artom | ;) | |
| 19:20:06 | artom | And that live migration is admin-only (by default)? | |
| 19:21:12 | cfriesen | artom: cold is too, isn't it? | |
| 19:22:02 | artom | cfriesen, doh, you're right | |
| 19:22:21 | imacdonn | to a typical sysadmin (I'm thinking private cloud), "migrate" means "VM is running on node A, and I want it to be running <somewhere else>" .. if we can do it without shutting the VM OS down, that'll make my life better | |
| 19:22:29 | artom | Wait no, I was looking at the wrong bit of api-ref | |
| 19:22:42 | imacdonn | from that perspective, they're basically the same thing | |
| 19:22:44 | artom | ... aaand no, still admin-only | |
| 19:23:41 | artom | I dunno, there's just a whole bunch of things that cold migration can do that live migration can't | |
| 19:24:20 | artom | And anyways, I was being rhetorical, we're obviously not going to limit osc migrate to cold migration and add a new osc live-migrate at this point. | |
| 19:27:27 | mriedem | cold migrate is admin only, resize is non-admin | |
| 19:27:32 | mriedem | w/ cold migrate under the covers | |
| 19:27:46 | artom | Kinky. | |
| 19:28:02 | mriedem | only on anniversaries dude | |
| 19:28:13 | mriedem | resize and cold migrate have been married awhile | |
| 19:28:59 | artom | Yeah, I'm surprised there's still anything going on under the covers. | |
| 19:29:08 | imacdonn | better than cold feet | |
| 20:07:24 | mriedem | melwitt: i think the vmware live migration change can go into a runway slot | |
| 20:07:34 | mriedem | rgerganov updated it to get ci passing, | |
| 20:07:41 | mriedem | i've done another pass, still -1 but it's closer | |
| 20:08:24 | melwitt | ok, cool. I'll add it then. missed the +1 vote from the vmware CI earlier today after I rechecked it | |
| 20:49:57 | melwitt | nova meeting in 11 minutes | |
| 20:50:06 | melwitt | 10 minutes | |
| 20:52:36 | mriedem | melwitt: want to hit this backport? https://review.openstack.org/#/c/605260/ - would be good to get those reverts in stable merged/released since they regressed blazar | |
| 20:52:50 | mriedem | the fix to replace the original is stacked on top on master but the gate is....not cooperating | |
| 20:53:53 | melwitt | can do. any idea why lee removed his vote? | |
| 20:57:36 | mriedem | no idea | |
| 20:57:38 | mriedem | classic lee | |
| 22:45:56 | cfriesen | is there an equivalent of CONF.reserved_huge_pages but for regular memory? (per numa node though) | |
| 22:46:58 | cfriesen | I can see CONF.reserved_host_memory_mb but that's not specifically 4k pages and is per compute node, not numa node. | |
| 22:47:44 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/pike: Update RequestSpec.flavor on resize_revert https://review.openstack.org/605879 | |
| 23:01:11 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/ocata: Update RequestSpec.flavor on resize_revert https://review.openstack.org/605880 | |
| 23:03:36 | mriedem | stable branch core review please https://review.openstack.org/#/c/600113/ | |
| 23:08:45 | cfriesen | mriedem: what would you think of something like CONF.reserved_huge_pages but for regular memory? (ie, to reserve specific amounts of 4K memory on each host numa node) | |
| 23:10:35 | cfriesen | or can CONF.reserved_huge_pages be used for 4k pages as well even though the name implies otherwise | |
| 23:11:14 | mriedem | cfriesen: i don't think you realize that i don't know anything about that nova/virt/hardware stuff | |