Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-02-09
19:39:27 mriedem psh
19:39:29 mriedem :P
19:39:31 cfriesen dansmith: private patches?
19:39:38 dansmith cfriesen: against the api? yeah :)
19:39:52 mriedem changing the api is interop sin
19:40:00 mriedem if you don't care about interop, then sin away
19:40:13 mriedem we all know canada is rife with sinners
19:40:19 dansmith and if you don't care about killing the rest of the world's interop bunnies
19:40:26 dansmith mriedem: true story
19:40:38 mriedem sinners that just can't get enough denim!
19:41:03 cfriesen dansmith: mriedem: yeah, I know. currently interop isn't a bit deal, the people that want the extra features are already locking themselves in. Personally I want to add an orthogonal vendor API microversion so that we would give vanilla behavior to vanilla clients. Havent' got resources for it though.
19:42:29 cfriesen mriedem: do you know if sdague ever had code for that Mitaka spec?
19:43:15 sdague cfriesen: I did not, we kept kind of going around in circles about the data stored in it
19:43:28 melwitt I had been thinking extra_specs showed on a 'nova show'. guess not
19:43:32 mriedem oomichi: on https://review.openstack.org/#/c/410006/ i don't see an existing test that handles the bdm boot_index schema
19:43:58 sdague because, after looking into it more in depth, it was actually going to be less API calls, less wire data, and less confusing to just include those attributes inline in the server response
19:44:36 cfriesen melwitt: sdague: yes, was wondering about that...just adding the extra-specs to the "nova show" output
19:44:38 sdague although, I guess if we had extra specs merged in it might get messy
19:44:55 sdague I don't know full, it needs someone to ponder the implications
19:44:59 mriedem plus aren't extra specs hidden from non-admin?
19:45:08 mriedem maybe that doesn't matter here
19:45:23 mriedem we're going to show them or not show them regardless of where we dump the flavor
19:45:23 sdague mriedem: it's fine if there are multiple visibility layers here I think
19:45:30 sdague right
19:45:34 melwitt could still make it admin on a nova show, if that's the case
19:45:38 mriedem yeah
19:45:51 mriedem i like to say things before i think
19:46:28 melwitt I can't remember if we ever updated 'nova show' to show the flavor stored on the instance instead of looking it up
19:46:41 mriedem melwitt: we don't look it up
19:46:47 melwitt *instead of looking it up in the flavors table I mean
19:46:56 mriedem we pull it from the instance
19:47:13 melwitt okay cool. so no more disappearing flavor on nova show after deleting a flavor
19:47:17 mriedem https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/views/servers.py#L248
19:47:29 cfriesen mriedem: are you suggesting that "nova flavor-list --extra-specs" doesn't work if you're not admin?
19:47:29 melwitt nice
19:47:45 melwitt okay, so yeah it seems like extra_specs could be added there, guarded by policy
19:48:09 melwitt (assuming it's already currently guarded by policy)
19:48:24 mriedem cfriesen: i guess i haven't tried
19:48:43 cfriesen mriedem: looks like maybe it's governed by policy: "compute_extension:flavorextraspecs:index": "rule:admin_or_owner"
19:49:03 mriedem and since only admins can create flavors by default, it'd be admin-only by default
19:49:38 cfriesen mriedem: I was wondering whether it might be readonly for regular users, to let them determine which flavors have the right extra-specs
19:50:12 cfriesen in our environment at least extraspecs create/update/delete is admin-only
19:51:05 artom Oh hey that's smart - if you only change the unit tests Jenkins doesn't all of CI, just the unit tests
19:51:18 mriedem artom: same for docs
19:51:47 artom mriedem, I knew about docs because dansmith mentioned it previously
19:54:38 openstackgerrit Anusha Unnam proposed openstack/nova master: Reno for additional-notification-fields-for-searchlight https://review.openstack.org/428835
19:56:18 Guest72572 Hi, I have a question about pci_whitelist_entry which are put in the nova.conf on the compute node.
19:57:00 openstackgerrit Ken'ichi Ohmichi proposed openstack/nova master: Remove unused validation code from block_device https://review.openstack.org/410006
19:57:10 oomichi mriedem: nice point, thanks. done on ^^^
19:57:15 Guest72572 is there a way / log which prints the compute node is correctly advertising its capabilities to the nova-scheduler that it has the desired number of VF's so that SRIOV ports can be created on this node ?
19:58:15 Guest72572 Cause, you get a generic error saying, "No Valid host found". So does any one know here if there are any logs which can help to debug this problem ?
19:59:46 cfriesen Guest72572: have you enabled debug logs in nova-scheduler?
20:07:22 Guest72572 cfriesen: Yes
20:07:55 Guest72572 The conductor itself will report that no valid host is found. Though I have setup sufficient VF's on the compute node.
20:13:06 cdent Guest72572: I seem to remember a bug that sounds a bit like this. What version are you?
20:13:15 Guest72572 @cfriesen: Is there any specific log that I should look for ?
20:14:19 Guest72572 Actually I'm using contrail distro. But the things is, I can see this for just one compute node. I works on other compute node.
20:14:54 Guest72572 *it works on other compute nodes.
20:15:34 Guest72572 So I'm trying to figure out, how to debug this , as in to find out if the compute node is correctly advertising these things or not.
20:16:57 cdent Guest72572: have you dug around in the n-cpu log for the failing compute node?
20:17:36 cdent it's responsible for auditing it's own resources
20:18:14 Guest72572 @cdent: Where can I find this log file ?
20:18:26 Guest72572 n-cpu log
20:18:27 cdent on the compute node itself
20:18:44 cdent what it is called and where it is depends a lot on how things were installed
20:18:54 cdent but if you've got a /var/log/nova, it's probably in there
20:19:50 Guest72572 sure
20:20:05 YmodeMonk Hi all. N00b question: I'm trying to list my compute service list, and from the compute node I get error: HTTP503 Server is termporarily unavailable.
20:20:13 diana_clarke Here are some manual testing notes against devstack (http://paste.openstack.org/show/598292/) that should make reviewing this easier: https://review.openstack.org/#/c/431714/
20:20:18 YmodeMonk (Setting a compute/controller node for the first time)
20:20:53 openstackgerrit Matt Riedemann proposed openstack/nova master: Use concurrency=eventlet in .coveragerc https://review.openstack.org/389824
20:22:43 openstack Launchpad bug 1663315 in cloud-init " identify openstack kvm platform on ppc64" [Medium,Confirmed]
20:22:43 mriedem mmedvede: fyi https://bugs.launchpad.net/nova/+bug/1663315
20:22:47 mriedem i'm not sure if that's a nova issue or not
20:23:21 openstackgerrit Artom Lifshitz proposed openstack/nova master: api sample tests: handle response missing an item https://review.openstack.org/431738
20:23:21 openstackgerrit Artom Lifshitz proposed openstack/nova master: Run api sample tests against 2.latest https://review.openstack.org/430352
20:30:02 mordred mriedem, sdague: I did a keystone equiv of the nova version discovery printing thing from yesterday. there are 4 different version discovery formats in the wild - so thanks for the nova documents all being the same
20:32:23 mordred cdent: ^^ you too http://paste.openstack.org/show/598293/ is the report - http://paste.openstack.org/show/598297/ has the raw json interspersed - the 'also' lines are because version endpoint returns a different payload and if we get it we take a stab at inferring an unversioned endpoint
20:33:20 mordred cdent: http://paste.openstack.org/show/598149/ is the nova one I pasted in yesterday if you didn't see that one
20:36:03 YmodeMonk Can anyone tell me why in my openstack catalog list I see "nova" and "compute" but no endpoint ? :-/ Any pointers as to what I may have done wrong or need to do?
20:37:56 mmedvede mriedem: that looks interesting, thanks for the tip!
20:38:52 YmodeMonk There are entries in the openstack service list, but openstack compute service list returns a 503 server currently unavailable.
20:39:25 YmodeMonk logs show database is locked.
20:39:36 YmodeMonk I think it's using sqlite3 instead of mysql :-/
20:42:00 sdague mordred: different version formats?
20:43:03 mordred sdague: the payload has differnet top level keys if it's from versioned or unversioned endpoint, and then some of the weirder clouds have different subkeys
20:43:17 sdague mordred: sigh :)
20:44:47 mordred sdague: so you can have dict(version={}), dict(versions=dict(values=[])), dict(versions=dict(version=[])) and dict(version=[]) - and in at least one case a thing with both values and version in the second level
20:45:22 mordred sorry, that last one should be dict(versions=[])
20:45:38 mordred sdague: yah. it's the hole I fell down today
20:47:46 cfriesen sdague: if we were to add flavor info to "nova show", how would you suggest formatting it? Add a dict to the existing "flavor" line? Or add a new property for the flavor dict?
20:49:59 openstackgerrit Diana Clarke proposed openstack/nova master: Fix nova-manage cell_v2 metavar strings https://review.openstack.org/431714
20:50:30 sdague cfriesen: well, for the basic attributes, I would put the keys directly inline I think. memory_mb, disk_gb, and the like
20:52:02 sdague for extra specs, I guess that would need a dedicated dict. I don't think we really pondered that here
20:52:27 cfriesen sdague: just include them as instance properties? what about extra_specs? call it "flavor_extra_specs" and put it inline? And what about odd ones like "rxtx_factor"
20:52:59 sdague cfriesen: honestly, this is basically where the whole debate is, what the representation looks like.
20:53:27 sdague honestly, if you want to get the ball rolling I would sketch out what the responses and flow would look like for a couple of different options
20:53:35 YmodeMonk In the [api_database] it still suggests that the connection = none ?
20:53:37 sdague fully inline, linked resource
20:53:57 sdague And get the vote on what's least gross :)

Earlier   Later