08:00:11 <licanwei> #startmeeting watcher
08:00:12 <openstack> Meeting started Wed Jun  5 08:00:11 2019 UTC and is due to finish in 60 minutes.  The chair is licanwei. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:00:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
08:00:15 <openstack> The meeting name has been set to 'watcher'
08:00:59 <chenke> say sum12: are you here?
08:01:19 <licanwei> Agenda for today: https://wiki.openstack.org/wiki/Watcher_Meeting_Agenda#06.2F05.2F2019
08:01:39 <Dantalion> sum12 is in the channel but not sure if afk or not
08:01:44 <licanwei> #topic Announcements
08:02:21 <licanwei> This week is Train-1 milestone.
08:03:02 <licanwei> We have completed 5 BPs
08:03:31 <licanwei> Thank you Dantalion and chenker!
08:03:50 <chenke> Thank you too.
08:04:06 <Dantalion> thanks everyone :)
08:04:13 <licanwei> #topic Review
08:04:16 <chenke> haha..
08:04:52 <licanwei> https://review.opendev.org/#/c/661245/ Add Placement helper needs review
08:05:30 <Dantalion> will have time to do reviews next week, placement helper is first priority (I am currently abroad)
08:05:45 <licanwei> Dantalion: thanks~
08:06:16 <chenke> where you are
08:06:28 <Dantalion> Back home in the Netherlands~
08:07:30 <licanwei> https://review.opendev.org/#/c/661121/ Optimize hypervisor API calls needs review
08:08:47 <licanwei> https://review.opendev.org/#/c/662089/ Optimize NovaHelper.get_compute_node_by_hostname needs review
08:09:12 <chenke> I think these patch are good.
08:09:34 <licanwei> They are Matt's commits
08:10:04 <licanwei> I'll review again~
08:10:34 <licanwei> https://review.opendev.org/#/c/661976/ Move datasource query_retry into baseclass needs review
08:11:09 <Dantalion> I need to update this patch to ensure backwards compatibility
08:11:32 <chenke> ok
08:13:27 <chenke> today i want to talk about instane.name and node.name
08:14:37 <licanwei> Dantalion: about backwards compatibility,
08:16:36 <licanwei> https://review.opendev.org/#/c/656622/
08:17:30 <licanwei> this merged patch has backwards compatibility problem
08:18:26 <Dantalion> Other than with strategy or datasource plugins?
08:19:06 <Dantalion> I think since the interface changed compatibility problems are impossible to prevent but happy to hear ideas
08:20:08 <licanwei> Dantalion: we changed the audit parameter,  about basic strategy , change node to compute_node , for example
08:21:15 <licanwei> In Stein, some tempest tests failed, https://review.opendev.org/#/c/662045/
08:21:19 <Dantalion> Ah yes that is true we could do that it can be both node and compute_node and stop support node in next release. I think compute_node is more clear but should support both for at least a while
08:22:04 <licanwei> Dantalion: yeh, I agree with you
08:22:18 <Dantalion> Do we want spec for this or should I just do it?
08:22:49 <licanwei> Dantalion: I don't think we need a spec, it's clear :)
08:23:03 <Dantalion> Okaj i'll start next week
08:23:18 <licanwei> Dantalion: nice~
08:23:49 <licanwei> https://review.opendev.org/#/c/649341/ Grafana proxy datasource to retrieve metrics needs new patch
08:24:15 <Dantalion> I need to rebase this to use new exception names
08:24:31 <licanwei> Dantalion: ok
08:24:52 <licanwei> https://review.opendev.org/#/c/661301/ Implement the configuration for Grafana datasource needs new patch
08:25:25 <Dantalion> I will add documentation as separate patch since it will be quite long to explain everything clearly
08:25:41 <chenke> good idea
08:25:48 <licanwei> Dantalion: thanks
08:26:02 <licanwei> https://review.opendev.org/#/c/660393/ Improve DevStack documentation to support metrics] needs review
08:27:22 <licanwei> I'll review again
08:28:00 <licanwei> last one:https://review.opendev.org/#/c/660011/ Add resource_name in action input parameter field needs new patch
08:29:28 <Dantalion> I will try to review next week
08:29:59 <Dantalion> after we have decided on humand_id / name
08:30:34 <licanwei> We need to define how to use humand_id /instance name/host name ...
08:30:56 <licanwei> # topic Discussion
08:31:22 <chenke> I think instance name is clear than human_id
08:31:29 <licanwei> Now let's discuss again
08:32:08 <licanwei> In Watcher data model, compute node and instance all have human_id field
08:32:24 <licanwei> example:
08:32:43 <licanwei> <ComputeNode status="enabled" disk_capacity="74" uuid="ubuntu-bionic-limestone-regionone-0007063015" hostname="ubuntu-bionic-limestone-regionone-0007063015" vcpus="8" human_id="" state="up" memory="7975" disk="73" disabled_reason="None" id="ae0b72a3-58b3-41d9-bd0d-5b765bfeb796">
08:32:58 <licanwei> <Instance disk_capacity="1" locked="False" uuid="3a2f9d3a-ccd0-4545-9556-291b7740df2b" watcher_exclude="False" vcpus="1" disk="1" state="active" memory="64" project_id="e24e6c9bbc6945c58da3bcc9c87d4929" human_id="tempest-testexecutestrategies-server-1452042406" metadata="{}"/>
08:33:34 <licanwei> Now computenode' s human_id is null
08:33:54 <licanwei> and compute node also has a hostname field
08:34:23 <licanwei> I think we should define how to use these fields
08:35:18 <chenke> I think uuid and hostname is enough.
08:35:34 <licanwei> remove human_id ,use hostname and instance name
08:35:53 <licanwei> or remove hostname and instance name, use human_id
08:35:56 <licanwei> ?
08:36:18 <licanwei> Now human_id is unused
08:37:16 <Dantalion> instance human_id is how we identify virtual machines for Grafana
08:38:07 <chenke> Why not using uuid?
08:38:08 <licanwei> Dantalion: Why don't you use vm uuid?
08:39:05 <Dantalion> Because storing hostname as identifier allows to query dns server to get ip without having to use openstack clients for example
08:39:39 <Dantalion> Which identifier to use for storing is not something I can control either is managed by other team
08:40:10 <Dantalion> but keep in mind what identifier to use is configurable and will differ in every deployment this is simply how CERN has done it.
08:41:00 <licanwei> Dantalion: Is VM name OK?
08:43:14 <chenke> let me see novaclient's code
08:44:38 <chenke> https://github.com/openstack/python-novaclient/blob/2595bac2294ce05e389eaab6636977963d5fc66c/novaclient/base.py#L152
08:44:59 <licanwei> Dantalion: Because instance human_id comes from nova client, maybe we plan to get data directely from nova API
08:46:15 <Dantalion> so human_id = instance_data['human_id'] ?
08:46:32 <licanwei> just like that placement helper
08:47:24 <Dantalion> Or do you mean if data from model is missing just use uuid to get object from nova API?
08:47:50 <chenke> human_id = instance_data['name']
08:48:48 <licanwei> Dantalion: No, I mean maybe we can get rid of novaclient, get data directely from nova API
08:50:17 <Dantalion> What is the benefit of using nova API directly?
08:52:38 <chenke> Now, when watcher create datamodel , instance's human_id actually is instance.name, but when updating datamodel, instance'human_id uses ['display_name']
08:53:47 <licanwei> Dantalion: If we use nova API directly, we will don't need Matt to change novaclient for optimization
08:53:59 <chenke> We need to confirm that their values are currently the same field
08:54:24 <chenke> I think we need novaclient.
08:54:30 <Dantalion> Ah I see, can someone show example difference between instance.name and display_name as I have no access to an openstack
08:54:59 <chenke> Let me have a try.
08:56:03 <licanwei> Dantalion: need to confirm the difference~
08:57:21 <licanwei> I'll confirm after the meeting
08:57:52 <Dantalion> Okaj yes I agree that the update should use same value as used when creating the data model.
08:58:16 <chenke> okaj
08:58:17 <licanwei> Dantalion: yes
08:58:31 <chenke> i will confirm this after our meeting
08:58:42 <Dantalion> We need to be careful to keep value that looks like hostname from compute_node in instance data model in the example: tempest-testexecutestrategies-server-1452042406
08:59:01 <Dantalion> this will be used at CERN to identify virtual machines in Grafana databases
08:59:22 <licanwei> Ok
08:59:27 <Dantalion> But I do not mind if the attribute is renamed
08:59:29 <licanwei> thank you all
08:59:38 <licanwei> I'll end the meeting
08:59:39 <Dantalion> Yes thank you
08:59:47 <chenke> ok
09:00:03 <licanwei> #endmeeting