Perl 6 - the future is here, just unevenly distributed

IRC log for #openstack-rally, 2016-05-24

| Channels | #openstack-rally index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:03 boomsauce joined #openstack-rally
00:14 yingjun joined #openstack-rally
00:19 diogogmt joined #openstack-rally
00:32 asti joined #openstack-rally
00:41 yanyanhu joined #openstack-rally
00:56 diogogmt joined #openstack-rally
01:14 akrzos joined #openstack-rally
01:49 ilbot3 joined #openstack-rally
01:49 Topic for #openstack-rally is now ☁ Rally RoadMap: http://goo.gl/JZkmwY ☁ Review Dashboard: http://goo.gl/YuCJfl ☁ IRC logs http://irclog.perlgeek.de/openstack-rally ☁ Key persons to ask:  boris-42, yfried, msdubov, rediskin, andreykurilin, amaretskiy  ☁ Documentation: https://rally.readthedocs.org/en/latest/ ☁ To publish changes to Rally:  https://rally.readthedocs.org/en/latest/contribute.html
01:51 openstackgerrit chenhb proposed openstack/rally: Add boot_and_delete_server samples and functional test  https://review.openstack.org/320180
02:22 MaxPC joined #openstack-rally
02:43 psuriset joined #openstack-rally
02:45 openstackgerrit chenhb proposed openstack/rally: Add boot_and_delete_server samples and functional test  https://review.openstack.org/320180
02:49 diogogmt joined #openstack-rally
02:52 arnoldje joined #openstack-rally
04:31 absubram joined #openstack-rally
04:32 absubram_ joined #openstack-rally
04:48 shausy joined #openstack-rally
05:02 bkopilov joined #openstack-rally
05:10 bkopilov joined #openstack-rally
05:11 bkopilov joined #openstack-rally
05:16 kambiz joined #openstack-rally
05:17 dmellado joined #openstack-rally
05:18 shausy joined #openstack-rally
05:32 rdas joined #openstack-rally
05:39 adiantum joined #openstack-rally
05:48 openstackgerrit SHIPeiqi proposed openstack/rally: CleanUp Error log traces in network context  https://review.openstack.org/319913
06:08 rcernin joined #openstack-rally
06:11 sverma joined #openstack-rally
06:17 meteorfox joined #openstack-rally
06:24 kun_huang joined #openstack-rally
06:28 nineseconds joined #openstack-rally
06:31 hughsaunders joined #openstack-rally
06:33 serverascode joined #openstack-rally
06:54 arcimboldo joined #openstack-rally
07:02 Deng joined #openstack-rally
07:04 tesseract joined #openstack-rally
07:21 arxcruz joined #openstack-rally
07:44 anshul joined #openstack-rally
07:49 openstackgerrit chenhb proposed openstack/rally: Fix list_samples unknown argument "limit" error  https://review.openstack.org/320264
07:52 ushkalim joined #openstack-rally
07:53 amaretskiy joined #openstack-rally
08:20 e0ne joined #openstack-rally
08:21 tosky joined #openstack-rally
08:26 tosky joined #openstack-rally
08:27 openstackgerrit Yanyan Hu proposed openstack/rally: Add Senlin plugin for profile and cluster  https://review.openstack.org/301522
08:32 e0ne joined #openstack-rally
08:40 ig0r_ joined #openstack-rally
08:41 ushkalim joined #openstack-rally
09:15 arcimboldo joined #openstack-rally
09:25 yanyanhu joined #openstack-rally
09:37 openstackgerrit Alexander Maretskiy proposed openstack/rally: [Install] Improve installation script  https://review.openstack.org/320324
09:44 arcimboldo_ joined #openstack-rally
09:45 arcimboldo_ Hi all, how can I see the output of the script run in a  VM.boot_runcommand_delete plugin?
09:51 ushkalim joined #openstack-rally
10:05 boomsauce joined #openstack-rally
10:11 MaxPC joined #openstack-rally
10:12 arcimboldo_ joined #openstack-rally
10:14 e0ne joined #openstack-rally
10:14 arcimboldo joined #openstack-rally
10:16 amaretskiy arcimboldo_: hi
10:16 amaretskiy you can generate html report
10:16 arcimboldo_ but I don't see anything in the report...
10:19 amaretskiy arcimboldo_: I've found a small bug - we have to modify standard plugin VM.boot_runcommand_delete in order to see results in html report because plugin return value is not ignored, instead plugin should call add_output() method
10:20 amaretskiy arcimboldo_: please look at https://github.com/openstack/rally/blob/master/rally/task/scenario.py#L172-L205
10:20 amaretskiy arcimboldo_:  you can use add_output() in any plugin
10:21 amaretskiy arcimboldo_: there are examples how output can be displayed: http://logs.openstack.org/42/318642/4/check/gate-rally-dsvm-rally/c60f41d/rally-plot/results.html.gz#/Dummy.dummy_output/output
10:21 amaretskiy arcimboldo_:  unfortunately we should update VM.boot_runcommand_delete a bit
10:22 amaretskiy arcimboldo_: I will do this update soon
10:22 mvk joined #openstack-rally
10:23 arcimboldo thnx amaretskiy, I'll wait for the update
10:28 openstackgerrit Sergey Skripnick proposed openstack/rally: [CI] Use --fromenv to create deployment on dsvm  https://review.openstack.org/318642
10:41 e0ne_ joined #openstack-rally
10:45 ushkalim joined #openstack-rally
10:46 openstackgerrit Antonio Messina proposed openstack/rally: Fix "AttributeError" when using "existing_users" context  https://review.openstack.org/320005
10:47 akrzos joined #openstack-rally
11:00 lazy_prince joined #openstack-rally
11:15 arcimboldo joined #openstack-rally
11:16 openstackgerrit Alexander Maretskiy proposed openstack/rally: [Plugins] Add deprecation mark to dummy scenario  https://review.openstack.org/320365
11:18 arcimboldo_ joined #openstack-rally
11:28 akrzos joined #openstack-rally
11:29 abalutoiu joined #openstack-rally
11:54 ushkalim joined #openstack-rally
11:56 ig0r_ joined #openstack-rally
11:58 nineseconds joined #openstack-rally
12:00 asti joined #openstack-rally
12:16 openstackgerrit Roman Vasilets proposed openstack/rally: Remove deprecated create_meter_and_get_stats  https://review.openstack.org/320403
12:30 maeca1 joined #openstack-rally
12:33 stpierre joined #openstack-rally
12:35 psuriset joined #openstack-rally
12:49 catintheroof joined #openstack-rally
12:53 diogogmt joined #openstack-rally
12:57 vponomaryov joined #openstack-rally
13:03 abalutoiu_ joined #openstack-rally
13:07 arnoldje joined #openstack-rally
13:24 nineseconds joined #openstack-rally
13:28 yingjun joined #openstack-rally
13:28 nineseconds joined #openstack-rally
13:38 openstackgerrit Alexander Maretskiy proposed openstack/rally: [Plugins] Fix custom output from boot_runcommand_delete()  https://review.openstack.org/320460
13:38 nineseconds joined #openstack-rally
13:42 yingjun joined #openstack-rally
13:48 amaretskiy arcimboldo: https://review.openstack.org/#/c/320460/
13:48 nineseconds joined #openstack-rally
13:54 psd joined #openstack-rally
13:59 yingjun joined #openstack-rally
14:09 yingjun joined #openstack-rally
14:10 openstackgerrit Merged openstack/rally: Fix "AttributeError" when using "existing_users" context  https://review.openstack.org/320005
14:13 openstackgerrit Alexander Maretskiy proposed openstack/rally: [Plugins] Fix custom output from boot_runcommand_delete()  https://review.openstack.org/320460
14:13 arcimboldo joined #openstack-rally
14:14 diogogmt joined #openstack-rally
14:14 KevinE joined #openstack-rally
14:15 KevinE joined #openstack-rally
14:15 _erick0zcr morning
14:18 openstackgerrit Alexander Maretskiy proposed openstack/rally: [Plugins] Add deprecation mark to dummy scenario  https://review.openstack.org/320365
14:18 arnoldje joined #openstack-rally
14:18 KevinE morning!
14:23 stpierre amaretskiy: so is it possible in https://review.openstack.org/#/c/320460/2/rally/plugins/openstack/scenarios/vm/vmtasks.py L196 for the data type to be a list or scalar? (e.g., in the case of an improperly written scenario)
14:25 amaretskiy stpierre: this can be any type that returned by json.loads (python simple types like dict, list, str, int, float)
14:25 amaretskiy stpierre: this can happen in case if command (this is usually a bash code given to scenario) returns data in unexpected format
14:26 stpierre ok, then let's log an error in that case
14:26 amaretskiy stpierre: we can improve this scenario a lot later, but let's just fix output for now
14:28 stpierre adding one line to ensure that we don't silently drop data is worth it now IMO
14:28 KevinE joined #openstack-rally
14:28 stpierre i'll happily +2 it as soon as that's added
14:30 logan- joined #openstack-rally
14:30 _erick0zcr stpierre: this is the error I had about http requests to a https service http://pastebin.com/rR8Ue2VX. For glance, I have a nginx doing reverse proxy, the openstack client works as expected.
14:34 e0ne joined #openstack-rally
14:39 yingjun joined #openstack-rally
14:42 stpierre _erick0zcr: requests is making an https request, so the error might be with your nginx proxy
14:42 stpierre but obviously if osc works that's a bit odd
14:43 stpierre does the curl command that requests logs on line 19 give you a 400?
14:43 stpierre this is suspicious: --cert None --key None
14:45 yingjun joined #openstack-rally
14:50 KevinE Guess I'll ask again: If anyone can help me try to troubleshoot gate-rally tests failing, I'd really appreciate it if you messaged me <3
14:51 stpierre KevinE: at this point, really the only course forward is to get devstack going and play with it there. i know you're new to openstack, but consider this a leanring opportunity with a very steep curve :)
14:53 openstackgerrit Merged openstack/rally: [Verify] Correctly configure auth URLs and auth version  https://review.openstack.org/319925
14:53 amaretskiy KevinE: hi
14:54 amaretskiy KevinE: please remind what is the patch?
14:54 KevinE stpierre: I'm trying my hardest I promise lol
14:55 KevinE amaretskiy: https://review.openstack.org/#/c/320056/ , more debug in the related change Disable osresources
14:55 _erick0zcr stpierre: omg, didn't notice the missing cert/key, lemme dig into
14:55 amaretskiy KevinE: I will take a look
14:59 rcernin joined #openstack-rally
15:01 amaretskiy KevinE: your change does not actually work, at least for keystone v2
15:01 amaretskiy KevinE: please point us to the proper doc/commit/etc which introduces parameter "interface"
15:02 amaretskiy KevinE: Also it is desired to mention such references in commit message
15:02 absubram joined #openstack-rally
15:03 yingjun joined #openstack-rally
15:04 wootehfoot joined #openstack-rally
15:05 KevinE amaretskiy: is this what you're referring to? https://review.openstack.org/#/c/318136/
15:06 amaretskiy KevinE: this is about deployment
15:06 amaretskiy KevinE:  however your patch changes osclients
15:08 amaretskiy KevinE: i guess there are proper changes in keystoneclient which expect this parameter
15:08 KevinE amaretskiy: is the catalog service supported for keystone v2?
15:09 amaretskiy KevinE: any cloud can provide keystone v2
15:10 arcimboldo joined #openstack-rally
15:10 diogogmt joined #openstack-rally
15:11 amaretskiy KevinE: I do not actually know if problem is related to keystone v2 only, the question is the reference to some docs about keystone API, or keystoneclient commits, etc which can explain that this parameter ("interface") is actually required
15:12 nineseconds joined #openstack-rally
15:18 e0ne joined #openstack-rally
15:20 openstackgerrit Merged openstack/rally: [Reports] Introduce class processing.plot.Trends  https://review.openstack.org/294125
15:25 _erick0zcr stpierre: the curl request made by rally uses glance v1 api, I don't know if it actually matters. Also the curl request output: curl: (35) Unknown SSL protocol error in connection to hyperclr.zpn.example.com.com:9292. Is there a way to tell rally to use certain api version? anyway I'll check the nginx side
15:26 stpierre it doesn't matter, but yes you can: https://github.com/openstack/rally/blob/master/rally-jobs/rally.yaml#L895-L897
15:27 _erick0zcr thanks
15:32 _erick0zcr omg, it actually works, it was the version that was messing it. Thanks again stpierre
15:40 vponomaryov joined #openstack-rally
15:40 asti joined #openstack-rally
15:40 ushkalim joined #openstack-rally
15:40 rmk joined #openstack-rally
15:41 Junhongl_ joined #openstack-rally
15:43 madhu_ak joined #openstack-rally
15:43 adiantum joined #openstack-rally
15:44 Junhongl_ joined #openstack-rally
15:45 arcimboldo joined #openstack-rally
15:45 mvk joined #openstack-rally
15:48 Junhongl joined #openstack-rally
15:59 diogogmt joined #openstack-rally
16:06 mvk joined #openstack-rally
16:08 KevinE_ joined #openstack-rally
16:09 lezbar joined #openstack-rally
16:22 lakerzhou joined #openstack-rally
16:30 openstackgerrit Roman Vasilets proposed openstack/rally: Remove deprecated create_meter_and_get_stats  https://review.openstack.org/320403
16:39 KevinE_ anyone know what this error is?:
16:39 KevinE_ WARNING keystoneclient.auth.identity.base [-] Failed to contact the endpoint at http://internalAddress:35357 for discovery. Fallback to using that endpoint as the base url.WARNING rally.common.broker [-] Failed to consume a task from the queue: Unable to establish connection to http://internalAddress:35357/domains/default
16:46 stpierre the machine running rally couldn't reach http://internalAddress:35357
16:51 harlowja joined #openstack-rally
16:54 KevinE_ but the endpoint_type is public, it should be going somewhere els
16:56 stpierre a) is your deployment using a public auth_url? b) does your catalog have the correct URLs for the public endpoints?
16:57 stpierre if you configure your deployment to use a public endpoint_type but with a private auth_url, for instance, then you could run into that
16:57 stpierre i think it's probably pretty normal to use the admin endpoint in the deployment, but ask for public URLs from the catalog, for instance
16:58 stpierre but if you can't hit the admin endpoint from your rally box, then you're dead in the water
16:59 e0ne joined #openstack-rally
17:09 KevinE_ how could I check for this?
17:10 KevinE_ and if that were wouldn't that mean that my patch wouldn't work regardless?
17:12 stpierre check the file from which your deployment was created
17:12 stpierre or 'rally deployment show <name>'
17:12 stpierre and also 'openstack catalog list' to check the catalog
17:22 KevinE_ json file: http://ip:adminport/v3, enpoint_type:public, it's failing to hit the admin endpoint
17:30 e0ne joined #openstack-rally
17:34 stpierre so it sounds like there's probably a firewall or something preventing rally from hitting that endpoint
17:57 KevinE_ so why would changing service_catalog.py or osclients fix it?
17:58 pvaneck joined #openstack-rally
17:59 adiantum joined #openstack-rally
17:59 dave-mccowan joined #openstack-rally
18:09 openstackgerrit Roman Vasilets proposed openstack/rally: [WIP] Refactor ceilometer plugin  https://review.openstack.org/320607
18:23 ig0r_ joined #openstack-rally
18:34 stpierre i have no idea why that would fix it
18:34 stpierre didn't you have endpoint_type in your deployment file twice at some point?
18:44 sverma joined #openstack-rally
18:50 KevinE_ yes, but someone (maybe you?) pointed out that the first one is just ignored due to the nature of json
18:51 stpierre i pointed out that one of them is ignored
18:51 stpierre i have no idea which one
18:52 stpierre there's one reasonable way to know for sure :)
18:52 stpierre only one of your two endpoint_type settings is actually correct, so just remove the other one
18:52 KevinE_ Just tried it, no difference
18:53 KevinE_ I think I'm going to submit another change (our service_catalog.py hotfix) just to see what jenkins thinks of it
18:57 KevinE_ nevermind
18:58 stpierre with your service_catalog.py fix, you should be able to run rally in debug mode and see exactly what requests it makes
18:58 stpierre if it's not making the admin endpoint request -- which we suspect it isn't -- then i guess dig into the code a bit more to try and figure out why
19:01 KevinE_ the fix pushes it through as admin and works flawlessly
19:02 KevinE_ the same with the osclients fix
19:02 stpierre so it hits the admin endpoint, but somehow magically works?
19:04 KevinE_ Oh sorry, it hits the public endpoint with the service fix
19:04 KevinE_ thus working lol
19:04 stpierre so, why not just configure your deployment to use the public endpoint?
19:05 KevinE_ I think I'm going to ask to be assigned to a different project this is hopeless
19:07 lakerzhou Hi. almost all nova test cannot specify network ID. It seems network ID is supported as an input to boot an instance from nova service plugin.
19:08 lakerzhou is there a workaround for the issue? because I cannot run any nova benchmark on production
19:10 stpierre you should be able to specify a network id to any scenario
19:10 stpierre they all support arbitrary keyword arguments to pass through to the underlying nova.servers.create() api call
19:12 lakerzhou @stpierre, how to use the arbitrary keyword? can you please give me an example? it has bother me for days.
19:14 stpierre https://paste.fedoraproject.org/370489/11724414/
19:14 stpierre look through the scenarios, they all support this
19:14 stpierre some of them do it in different ways -- boot_server_from_volume_and_resize takes separate boot_server_kwargs and create_volume_kwargs, for example
19:14 stpierre but anything you can do in the API call can be passed to Rally
19:16 lakerzhou thanks a lot! I will give it a try.
19:22 lakerzhou I am trying to develop a tool record historical benchmarks and use the data to inject SLA to new test. For example, if the execution time exceed two times of the average max duration, the test will be fail.
19:23 lakerzhou is there any build-in rally tools supporting this kind of usage?
19:24 lakerzhou of course, the historical data will be stored for each benchmark test case.
19:25 stpierre not that i know of, but that would be a nice SLA plugin to have
19:27 mvk joined #openstack-rally
19:27 lakerzhou I am wondering what the rally DB used for? Seems not much documents for it.
19:29 stpierre it tracks all of the data -- deployments, tasks, verifications, and so on
19:32 lakerzhou ok, I see. I guess there is not rest api support for the DB access.
19:33 stpierre there's no rest api at all yet :)
19:33 stpierre but you should be able to make an SLA plugin easily enough
19:33 stpierre and i think it sounds awesome, so i'd love to have it upstream
19:35 lakerzhou I am not familiar the whole "plugin" architecture yet. But it is supposed to be my task for the next few weeks.
19:36 lakerzhou my main concern now is where/how to store the historical data
19:36 stpierre it's already stored in the db
19:37 lakerzhou really, that is much handy than I thought.
19:38 e0ne joined #openstack-rally
19:39 adiantum joined #openstack-rally
19:45 KevinE_ So where I’m at right now: two of the top people in the IRC helped me develop a change that will not merge. one person says that it's our firewall, someone else says that it doesn't work for keystone v2, someone else says that the change doesn't work at all, and someone else says that the change works great, and no one knows why the tests are failing to merge. What am I supposed to do
19:50 stpierre reconfigure your deployment to hit the public interface instead of admin
19:50 stpierre and/or figure out why you can't hit the admin interface
19:50 wootehfoot joined #openstack-rally
19:51 stpierre both of those things are easily testable and you can start crossing things off the list
19:52 stpierre as far as we can tell, your change just makes rally hit the public interface instead of the admin interface, and there's a way to do that without changing code. so, try it.
20:02 KevinE_ reconfigured the deployment to hit the public interface instead of admin, did not work, still attempted to contact the admin endpoint for discovery and failed
20:04 stpierre that seems wrong
20:08 stpierre i guess not, i can replicate that
20:09 KevinE_ http://pastebin.com/xDvSr2rH
20:12 KevinE_ ok so if it can be replicated, how can it be solved that's not the path I already tood
20:12 KevinE_ took *
20:17 stpierre so, your change forces rally to avoid the admin interface
20:17 stpierre but the problem that the gate runs into (and that i'm hitting as well) is that trying to create a tenant via the public interface gives a 404
20:18 stpierre i don't know enough about keystone to know if there's an elegant solution to that, but it seems to me that we need the admin interface, at least for tenant creation
20:19 stpierre if your fix works at your site, then clearly there must be some kind of workaround to allow tenant creation via the public interface
20:19 stpierre but what you added should, at best, be an option, not the default
20:19 KevinE_ rally automatically decides that it wants the interface to be admin. My change just tells it to use what the user puts in
20:20 stpierre not rally, keystoneclient
20:20 stpierre and it decides that because it knows (or thinks) that it has to use the admin interface to create a tenant
20:21 stpierre which is true, at least in some (or most?) installations
20:21 stpierre i don't know why it's apparently not true in yours
20:22 stpierre IMO the easiest solution here would be to run rally from somewhere that can hit the admin interface
20:22 stpierre failing that, it seems like adding another option to the deployment config to specify the interface might work, but clearly we can't pin interface == endpoint_type
20:23 stpierre or, you could run rally with existing users so that it doesn't try to create tenants, but that seems like the nuclear option
20:28 KevinE_ we don't restrict admin rights at this point
20:29 KevinE_ under our Icehouse implementation, we have a "repose" server sitting in front of each API
20:29 KevinE_ the repose server block remote admin capabilities
20:29 KevinE_ ^that lol
20:30 stpierre oh, then i suppose you'll have no choice but to try to run rally with existing users
20:30 stpierre or run it from somewhere where it doesn't have to go through repose
20:34 KevinE_ ok existing users is different
20:35 KevinE_ for Icehouse, we are using a non-Openstack identify management setup
20:35 KevinE_ so, the ability for Rally to create/delete/modify users or tenants is blocked
20:35 KevinE_ for the Liberty deployment, we are using local keystone user and project management - which gives us the ability to use either dynamic or static users
20:36 KevinE_ stpierre: <for the tag
20:40 stpierre for icehouse you'll have no choice but to use existing users. rally still requires admin for some things, so it still might not work -- you'll just have to TIAS
20:44 stpierre for liberty, it'll depend on whether or not you still use the repose gateway
20:50 e0ne joined #openstack-rally
20:52 dave-mccowan hello rally gurus... does anyone have experience working with Rally, Heat, and TLS?
20:53 dave-mccowan When running heat or openstack cli, my tests run fine.  But when running with Rally and Heat together, I am getting an SSL certificate error.  Has anyone seen this before?
21:02 rcernin joined #openstack-rally
21:04 KevinE_ why does the openstack CLI differentiate between endpoint_type and OS-interface, but rally does not
21:06 KevinE_ is an openstack cli interface = to a rally endpoint_type?
21:12 KevinE_ stpierre: if you could answer those questions, and then I think that setting my change to be optional would be the way to go. I feel like I understand where you're coming from with that now
21:12 stpierre i don't know the answer to either of them
21:12 stpierre sorry
21:13 KevinE_ lol
21:13 KevinE_ well, how about how do I make progress in making that some sort of an optional parameter
21:14 stpierre add it to https://github.com/openstack/rally/blob/master/rally/deployment/engines/existing.py
21:14 stpierre i haven't done much with the deployments so i'm not sure exactly how all of that information gets passed to osclients
21:14 stpierre you're in for some RTFS
21:16 KevinE_ i'm scared to ask what that acronym means
21:17 stpierre read the fine source
21:17 stpierre or some other f-word if you prefer
21:31 MaxPC joined #openstack-rally
21:37 diogogmt joined #openstack-rally
21:43 KevinE_ so you think that my specific edit will get merged upstream to be one of the basic parameters used to create a rally deployment? with how terribly this has been going so far I can't see that really happening?
21:55 KevinE joined #openstack-rally
22:48 lakerzhou joined #openstack-rally
22:51 akrzos joined #openstack-rally
23:03 openstackgerrit joined #openstack-rally
23:05 zackf joined #openstack-rally
23:05 adiantum joined #openstack-rally
23:26 akrzos joined #openstack-rally
23:51 akrzos joined #openstack-rally

| Channels | #openstack-rally index | Today | | Search | Google Search | Plain-Text | summary