Perl 6 - the future is here, just unevenly distributed

IRC log for #openstack-rally, 2015-04-15

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

All times shown according to UTC.

Time Nick Message
00:35 dmorita joined #openstack-rally
00:38 baker joined #openstack-rally
00:48 rook_ joined #openstack-rally
00:50 rook_ joined #openstack-rally
01:12 openstackgerrit Kun Huang proposed stackforge/rally: add test for deployment check  https://review.openstack.org/170342
01:18 prashantS joined #openstack-rally
01:27 prashantS joined #openstack-rally
01:49 rook_ joined #openstack-rally
02:35 openstackgerrit Merged stackforge/rally: Data structs declared with constructor replaced with literal  https://review.openstack.org/173410
02:57 rook_ joined #openstack-rally
02:59 rook___ joined #openstack-rally
03:25 aswadr joined #openstack-rally
03:25 sputnik13 joined #openstack-rally
03:29 prashantS joined #openstack-rally
03:40 prashantS joined #openstack-rally
04:01 rook_ joined #openstack-rally
04:03 rook___ joined #openstack-rally
04:04 rook_ joined #openstack-rally
04:15 pradeep joined #openstack-rally
04:22 rook___ joined #openstack-rally
04:23 pradeep1 joined #openstack-rally
04:30 davideagnello joined #openstack-rally
04:57 rdas joined #openstack-rally
05:00 nkhare joined #openstack-rally
05:03 rook_ joined #openstack-rally
05:14 tfreger joined #openstack-rally
05:40 sputnik13 joined #openstack-rally
05:59 bkopilov joined #openstack-rally
06:03 neeti joined #openstack-rally
06:04 rook_ joined #openstack-rally
06:06 rook___ joined #openstack-rally
06:15 rdas joined #openstack-rally
06:33 kiran joined #openstack-rally
06:36 kiran-r pradeep1: Do I have to write unit tests to every method I am committing?
06:36 pradeep1 kiran-r: preferably yes.
06:37 pradeep1 kiran-r: if not some one will ask you to do :)
06:37 pradeep1 kiran-r: what are you writing?
06:37 kiran-r pradeep1: Thanks! :)
06:50 bkopilov joined #openstack-rally
06:52 openstackgerrit Pradeep K Surisetty proposed stackforge/rally: Add Cinder create backup, list backup scenario  https://review.openstack.org/173482
06:57 sputnik1_ joined #openstack-rally
07:00 sputnik13 joined #openstack-rally
07:02 sputnik__ joined #openstack-rally
07:05 rook_ joined #openstack-rally
07:07 jeblair joined #openstack-rally
07:11 bkopilov joined #openstack-rally
07:31 sputnik13 joined #openstack-rally
07:47 fhubik joined #openstack-rally
07:49 bkopilov joined #openstack-rally
07:50 arxcruz joined #openstack-rally
07:54 karimb joined #openstack-rally
07:54 bkopilov joined #openstack-rally
08:02 e0ne joined #openstack-rally
08:06 rook_ joined #openstack-rally
08:08 rook___ joined #openstack-rally
08:08 e0ne joined #openstack-rally
08:08 davideagnello joined #openstack-rally
08:17 bkopilov joined #openstack-rally
08:23 e0ne joined #openstack-rally
08:31 openstackgerrit Pradeep K Surisetty proposed stackforge/rally: Add Cinder create backup, list backup scenario  https://review.openstack.org/173482
08:33 e0ne joined #openstack-rally
08:36 cdent joined #openstack-rally
08:37 openstackgerrit Yuriy Nesenenko proposed stackforge/rally: Add Cinder restore_volume_backup scenario  https://review.openstack.org/173412
08:40 zerda joined #openstack-rally
08:45 bkopilov joined #openstack-rally
08:52 bkopilov joined #openstack-rally
09:08 rook_ joined #openstack-rally
09:10 rook___ joined #openstack-rally
09:19 psd joined #openstack-rally
09:33 openstackgerrit Sergey Skripnick proposed stackforge/rally: WIP: super new rally_gate.py  https://review.openstack.org/163785
09:34 e0ne joined #openstack-rally
09:38 boris-42 yfried: ping
09:38 boris-42 yfried: I am going to start dicussion in mailing list regarding to Rally meeting
09:38 yfried boris-42: sure
09:38 boris-42 like we can move them to openstack-rally
09:39 boris-42 and chose any time
09:39 yfried boris-42: sure
09:44 boris-42 yfried: as well we would like to have 2 meeting / week
09:44 boris-42 yfried: 1 is for release management stuff
09:44 yfried boris-42: yeah. that was said on last night's mtg
09:44 boris-42 like what patches should be merged in what release
09:45 boris-42 so I think having dedicated person for all release management stuff is curcial if we would like to have release / 2 weeks
09:45 yfried boris-42: msdubov said he would own this
09:49 boris-42 yfried: yep
09:49 boris-42 yfried: I think he is the best person for doing such work
09:50 boris-42 yfried: at least better than me=)
09:50 rook_ joined #openstack-rally
09:59 tosky joined #openstack-rally
10:03 e0ne joined #openstack-rally
10:03 openstackgerrit Nikita Konovalov proposed stackforge/rally: Wait for status  https://review.openstack.org/172466
10:04 tfreger1 joined #openstack-rally
10:14 openstackgerrit Sergey Skripnick proposed stackforge/rally: WIP: super new rally_gate.py  https://review.openstack.org/163785
10:33 pradeep1 boris-42: 2 per week fine. But adjust time please
10:34 pradeep1 Regular time is mid night for me.
10:34 pradeep1 :)
10:43 boris-42 pradeep1: in our case it doesn't matter
10:43 boris-42 pradeep1: we are keeping master working so we don't have feature freeze / code freeze / qa process and so on for relases
10:45 pradeep1 pradeep1: ok
10:54 e0ne joined #openstack-rally
10:54 fhubik joined #openstack-rally
11:01 arxcruz joined #openstack-rally
11:04 pbandzi joined #openstack-rally
11:06 openstackgerrit Nikita Konovalov proposed stackforge/rally: [Sahara] More debug logging  https://review.openstack.org/170491
11:07 bkopilov joined #openstack-rally
11:19 openstackgerrit Pavel Boldin proposed stackforge/rally: Refactor run_command_over_ssh, add script_args  https://review.openstack.org/173371
11:38 e0ne joined #openstack-rally
11:45 dulek joined #openstack-rally
11:46 davideagnello joined #openstack-rally
11:47 dulek Hi, I've wanted to run NovaServers.boot_and_live_migrate_server on a VM that has some memory load on it. Is it possible to delay live migration action here to let the VM start to use memory?
11:48 dulek I've tried to start my stressing script as fast as possible, but even placing it in the beginning of a runlevel is too early
11:51 panbalag joined #openstack-rally
11:52 pradeep1 dulek: you want time between boot & migrate?
11:52 pradeep1 dulek:  you have too many dirty pages on VM, so you want it to be delayed or how is that?
11:53 pradeep1 dulek:  sleep in between
11:53 pradeep1 server = self._boot_server(image, flavor, **kwargs)
11:53 pradeep1 new_host = self._find_host_to_migrate(server)
11:53 pradeep1 self._live_migrate(server, new_host,
11:53 pradeep1 block_migration, disk_over_commit)
11:53 samueldmq left #openstack-rally
11:57 dulek Oh, I see, I can change scenario's code. :)
11:57 dulek pradeep1: Thanks!
11:59 boris-42 dulek: yep
11:59 boris-42 pradeep1: thanks
11:59 boris-42 dulek: there is speicall stuff sleep_between()
12:00 boris-42 dulek: https://github.com/stackforge/rally/blob/master/rally/benchmark/scenarios/nova/servers.py#L103
12:00 boris-42 dulek: so you can add it in the same way to live_migrate scenario
12:00 boris-42 dulek: I believe that can be even merged to upstream
12:02 dulek boris-42: I'll test it and take a look on that. :)
12:09 mwagner_lap joined #openstack-rally
12:17 boris-42 dulek: ok =)
12:26 fhubik joined #openstack-rally
12:28 aix joined #openstack-rally
12:30 e0ne joined #openstack-rally
12:31 openstackgerrit svasheka proposed stackforge/rally: Add keystone benchmark scenarios for roles  https://review.openstack.org/165409
12:37 openstackgerrit svasheka proposed stackforge/rally: Add keystone update_user_password scenario  https://review.openstack.org/165422
12:41 openstackgerrit svasheka proposed stackforge/rally: Add keystone create_update_and_delete_tenant scenario  https://review.openstack.org/165450
12:50 openstackgerrit Yuriy Nesenenko proposed stackforge/rally: Add Cinder create_backup volume scenario  https://review.openstack.org/167171
12:57 jaypipes joined #openstack-rally
12:58 baker joined #openstack-rally
13:41 yfried joined #openstack-rally
13:47 openstackgerrit Sergey Skripnick proposed stackforge/rally: WIP: super new rally_gate.py  https://review.openstack.org/163785
13:58 prashantS joined #openstack-rally
14:03 baker joined #openstack-rally
14:03 kiran-r Is there a way to create a new user and tenant every ¨time¨??
14:09 baker_ joined #openstack-rally
14:17 pradeep1 kiran-r: every time as in?
14:17 prashantS joined #openstack-rally
14:18 kiran-r Like every iteration, I want to create a new set tenant/users
14:18 pradeep1 kiran-r:  create_tenant_with_users in rally/benchmark/scenarios/keystone/basic.py
14:19 pradeep1 ir does
14:19 kiran-r I am sorry. I am unable to follow
14:20 kiran-r in an existing testcase
14:21 kiran-r say like create_and_list_networks, here it creates users/tenants only during initialization, after that, it simply populates
14:21 kiran-r Getting it?
14:42 openstackgerrit Cyril Roelandt proposed stackforge/rally: task report: generate a JUnit report  https://review.openstack.org/171664
14:44 baker joined #openstack-rally
14:54 neeti joined #openstack-rally
14:56 openstackgerrit svasheka proposed stackforge/rally: Add keystone create_update_and_delete_tenant scenario  https://review.openstack.org/165450
14:59 openstackgerrit Sergey Skripnick proposed stackforge/rally: WIP: super new rally_gate.py  https://review.openstack.org/163785
15:06 openstackgerrit svasheka proposed stackforge/rally: Add keystone benchmark scenarios for roles  https://review.openstack.org/165409
15:20 yingjun joined #openstack-rally
15:30 rdas joined #openstack-rally
15:32 openstackgerrit Yuriy Nesenenko proposed stackforge/rally: Add Cinder restore_volume_backup scenario  https://review.openstack.org/173412
15:33 baker_ joined #openstack-rally
15:37 prashantS joined #openstack-rally
15:48 csoukup joined #openstack-rally
15:53 cdent joined #openstack-rally
16:09 yfried boris-42: I had a strange though
16:09 yfried t
16:10 yfried boris-42: could we add a "no-cleanup" context/flag?
16:11 yfried boris-42: good for debug
16:13 pradeep joined #openstack-rally
16:14 kiran joined #openstack-rally
16:16 boris-42 yfried: so actually I was wating for garyk
16:16 boris-42 yfried: but seems like he is not able to do that
16:16 boris-42 yfried: so be free to add it
16:17 boris-42 yfried: it should be just few lines patch
16:21 tosky joined #openstack-rally
16:30 openstackgerrit Pradeep K Surisetty proposed stackforge/rally: Add Cinder create backup, list backup scenario  https://review.openstack.org/173482
16:31 openstackgerrit Yuriy Nesenenko proposed stackforge/rally: Add Cinder restore_volume_backup scenario  https://review.openstack.org/173412
16:31 openstackgerrit Yuriy Nesenenko proposed stackforge/rally: Add Cinder create_backup volume scenario  https://review.openstack.org/167171
16:31 pradeep guess dvsm gates are failing.
16:38 openstackgerrit Yuriy Nesenenko proposed stackforge/rally: Add Cinder restore_volume_backup scenario  https://review.openstack.org/173412
16:47 afazekas joined #openstack-rally
16:50 meteorfox boris-42: hey, I have a question. So, I have a few scenarios in YAML that I'm writing for Rackspace private cloud, I would like to share them with the community, do you have suggestion on how to include them in the Rally repo? The reason I asked, is because I don't think the samples/ directory is a good place for it, since it will probably look more like the
16:50 meteorfox stuff you have under rally-jobs/
16:50 boris-42 meteorfox: I will make soon blogpost
16:51 boris-42 meteorfox: it will be called cloud certification
16:52 boris-42 meteorfox: short idea is to have one parametrized template
16:52 meteorfox boris-42: ok cool. I was thinking of writing templated tasks, and then used task-args-file to feed in Rackspace specific scenario data
16:52 boris-42 meteorfox: that can certify cloud
16:53 boris-42 meteorfox: yep
16:53 boris-42 meteorfox: I believe you can add certification/ directory
16:53 boris-42 meteorfox: and send to review
16:55 meteorfox boris-42: awesome. Just the empty directory?
16:55 boris-42 meteorfox: with file that you have
16:57 pradeep looks like dsvm gates failing
16:57 pradeep known issue?
16:58 meteorfox boris-42: ok cool. Right now, I've only been parametrizing runner total, concurrency, and the image. My scenario is  not fully written yet, but I'll submit the patch to review, if I take too long just go ahead, and open it
17:00 meteorfox boris-42: also, I'll be looking forward for your blog post, let me know once is up :)
17:13 boris-42 meteorfox: ok
17:15 meteorfox boris-42: is there a context where I can say "take 100 image snapshots", then execute GlanceImages.list_images ?
17:16 boris-42 meteorfox: nope
17:16 meteorfox boris-42: ok
17:16 boris-42 meteorfox: but you can upload any amount of images
17:17 meteorfox boris-42: yeah, I have one scenario like that
17:17 boris-42 meteorfox: or use image-create-and-list benchmakr scenario
17:17 boris-42 meteorfox: I mean there is context for that
17:18 meteorfox boris-42: ok, that could work.
17:27 meteorfox boris-42: here's a description of the stuff I'm writing scenarios for, I need to contribute the cluster scenario. But most of them can be accomplished with rally. https://gist.github.com/meteorfox/f62d8e3ba7775722d2f5
17:28 davideagnello joined #openstack-rally
17:30 bkopilov joined #openstack-rally
17:32 prashantS joined #openstack-rally
17:33 meteorfox boris-42: those are just the basics. Later I plan to add swift, and other stuff
17:34 boris-42 meteorfox: so the idea is that we should do some math
17:34 boris-42 meteorfox: like
17:34 boris-42 meteorfox: amount of controller amount of computes amount of active users -> (concurrencty and times)
17:37 e0ne joined #openstack-rally
17:38 meteorfox boris-42: I understand what you are trying to do, but it might not work for all architectures (flavor definitions, resources per host and per controller, network bandwidth, image sizes, etc)
17:39 bkopilov joined #openstack-rally
17:40 meteorfox boris-42: perhaps we could have certain 'profiles', like small, medium, large, x-large clouds, where small is like 1 controller, 5 nodes, medium  is like 3 controllers, and 10 nodes, etc
17:41 meteorfox but again, 1 node of 120 GB ram, and 32 CPUs, != 1 node of  64GB and 16 CPUs
17:41 meteorfox so it might be challenging
17:50 meteorfox boris-42: also, maybe first we can come up with what we consider "typical" used flavors, and images sizes, tenants etc, then come up with performance criteria. Say, typical flavor is m1.medium, and typical image size, is 700 MB for base images, and snapshots are around 2GB. Then we aggregate all the resources from the hypervisors, and compute expected number
17:50 meteorfox of builds. Now for build times, it will be tricky, since it will depend on your network bandwidth, your storage backend, hypervisor, etc, Personally I'd say for the flavor/image combination above, it should be more than 1minute, but that might vary from architecture to architecture
17:52 meteorfox boris-42: but the point is that it just depends
17:54 boris-42 meteorfox: baby steps please
17:54 boris-42 meteorfox: let's start from enterprise case
17:55 boris-42 meteorfox: you have some amount of hardware and some amount of input users with some particaluar workflows
17:55 boris-42 meteorfox: you would like to check how it works and does it works well
17:55 yfried boris-42: still waiting for this review https://review.openstack.org/#/c/172831/ could you please take a look?
17:55 meteorfox right
17:58 andreykurilin joined #openstack-rally
18:07 meteorfox boris-42: heh. 503 from gerrit
18:08 prashantS joined #openstack-rally
18:09 boris-42 meteorfox: uh oh
18:09 meteorfox boris-42: are you also getting 503s?
18:10 meteorfox yfried: ^^
18:11 boris-42 meteorfox: I think they are updating gerrit
18:11 boris-42 meteorfox: at least I saw such patch in infra
18:11 boris-42 yfried: I will try
18:11 meteorfox boris-42: I thought that was going to happen on Friday, http://lists.openstack.org/pipermail/openstack-dev/2015-April/061489.html
18:12 boris-42 meteorfox: I saw this Change in openstack-infra/system-config[master]: Upgrade production gerrit to 2.8.4.17
18:14 meteorfox boris-42: heh, ok. I guess I should grab something to eat now then. I was reviewing yfried's spec
18:28 prashant_ joined #openstack-rally
18:36 prashantS joined #openstack-rally
18:39 davideagnello boris-42: hello, the Scenario class defined in base.py mentions that you can specify custom scenarios in test config.  Where do I find more information on the test config file?
18:40 andreykurilin joined #openstack-rally
18:41 boris-42 davideagnello: you mean rally input format?
18:43 boris-42 davideagnello: https://rally.readthedocs.org/en/latest/tutorial/step_2_running_multple_benchmarks_in_a_single_task.html ?
18:44 davideagnello boris-42: the comment in the scenario base class says: And your test scenarios will be auto discoverable and you will be able to specify it in test config.
18:44 boris-42 davideagnello: yep
18:45 davideagnello boris-42: the comment probably means the yaml or json file?
18:45 boris-42 davideagnello: yep
18:45 davideagnello boris-42: ok cool
18:46 boris-42 davideagnello: so you are just inherit that class and new benchmark scenarios are auto discovered
18:46 boris-42 davideagnello: if module is loaded
18:46 boris-42 davideagnello: so you can put code in ~./rally/plugins and it will work
18:46 davideagnello boris-42: yup, I got my custom scenario test running this way :)
18:46 boris-42 davideagnello: yep this is proper way to do things
18:47 davideagnello boris-42: trying to understand how results are being propagated up to rally framework
18:47 boris-42 davideagnello: so let me help you
18:47 oanufriev joined #openstack-rally
18:47 boris-42 davideagnello: this code executes task https://github.com/stackforge/rally/blob/master/rally/benchmark/engine.py#L209-L247
18:47 boris-42 davideagnello: so it is some kind of entry point*
18:49 davideagnello boris-42: ok
18:50 davideagnello boris-42: so results are contained in this results dict for each test run
18:50 boris-42 davideagnello: and then stored in DB
18:51 davideagnello ok, from the test scenario.  What can you constitute as results that would be captured by the framework?
18:52 davideagnello boris:  anything returned by a client the test is interacting with?  Can extra data that parses responses from clients be added too?  I have noticed some tests make use of assert
18:57 boris-42 davideagnello: so for now it will collect duration of single iteration (1 run of scenario)
18:57 boris-42 davideagnello: all atomic actions
18:57 boris-42 davideagnello: and data that was returned by scenario method
18:58 boris-42 davideagnello: for now returned data has very simple structure {key: number}
18:58 boris-42 davideagnello: in future it will be more powerful
18:58 davideagnello boris-42: returned data in a dictionary key/value pairs?
18:58 davideagnello boris-42: ok
18:59 boris-42 davideagnello: yep exactly like {"throughput": 10.5}
19:02 davideagnello boris-42: ok neat.  so this returned data can be used in the generated reports?
19:02 boris-42 davideagnello: yep it will
19:02 boris-42 davideagnello: one sec
19:02 baker joined #openstack-rally
19:02 boris-42 davideagnello: http://logs.openstack.org/88/164888/9/check/gate-rally-dsvm-rally/ca266f0/rally-plot/results.html.gz#/VMTasks.boot_runcommand_delete/output
19:02 boris-42 davideagnello: you'll get something like this
19:04 davideagnello boris-42: ok cool, this data is used to get a representation of what your scenario is doing
19:04 boris-42 davideagnello: yep it's usefull in case when you would like for example to benchmark some other system
19:05 baker joined #openstack-rally
19:05 boris-42 davideagnello: like analyzing VM performance
19:05 davideagnello boris-42: ok, makes sense
19:21 prashantS joined #openstack-rally
19:21 jaypipes joined #openstack-rally
19:30 e0ne joined #openstack-rally
19:48 prashantS joined #openstack-rally
19:54 bkopilov joined #openstack-rally
20:01 rook_ joined #openstack-rally
20:07 baker_ joined #openstack-rally
20:35 rook___ joined #openstack-rally
20:41 meteorfox boris-42: is there support already for concurrent scenarios?
20:42 boris-42 meteorfox: nope
20:42 boris-42 meteorfox: I hope it will be ready in half year=)
20:42 meteorfox boris-42: ah ok, right there was the spec.
20:42 boris-42 meteorfox: yep
20:43 boris-42 meteorfox: so actually what I think is to split rally into 2 repos
20:43 boris-42 meteorfox: rally and rally-openstack-plugins
20:43 meteorfox boris-42: ok. hmm.. what will the plugins hold?
20:44 boris-42 meteorfox: I would like to make rally indpeendet from openstack
20:44 boris-42 meteorfox: so it will be simpler to understand how it works
20:44 meteorfox boris-42: ooh
20:44 boris-42 meteorfox: less code
20:44 boris-42 meteorfox: just engine
20:45 boris-42 meteorfox: plus it will be simpler to use rally for your personal stuff
20:46 boris-42 meteorfox: it's not so hard task
20:46 meteorfox boris-42: hmm.. I understand, but in the past projects that have done this typically end up with one usage, and the abstraction just make things more complicated
20:47 meteorfox boris-42: for example, OpenCafe and CloudCafe
20:47 davideagnello boris-42: I have a scenario tests in which I am trying to return a dict of key/pairs back.  When I run my scenario, I get a failed validation because additional_properties are not allowed.  Base scenario class defines this as False.  Does it make sense for a custom scenario to override this CONFIG_SCHEMA value and set it to true?
20:47 boris-42 meteorfox: we are in different situation
20:52 meteorfox boris-42:  flesh it out in a spec, and let's discuss it. I'm not opposed, but right now I don't see a big benefit, and in my mind Rally will probably continue just to be an OpenStack framework. Unless you are already using some portions of Rally, and see a benefit in re-using some of its core
20:53 boris-42 meteorfox: I mean it can't be implmeneted now
20:53 meteorfox :)
20:53 boris-42 meteorfox: there are issues
20:53 boris-42 meteorfox: so we are in different situation because I made very-specific-for-openstack-benchmark-tool
20:53 boris-42 meteorfox: and then I start working on generalization
20:53 meteorfox boris-42: ok, I see, so you want to start decoupling it
20:53 boris-42 meteorfox: with keeping (or even making it simpler)
20:54 boris-42 meteorfox: so the idea is nex
20:54 boris-42 meteorfox: to make all openstack related stuff in openstack
20:54 boris-42 meteorfox: keeping simplicity
20:55 boris-42 meteorfox: after that just move (plugins + requirements)  in separeted project
20:55 boris-42 meteorfox: after that rally will become more lightweight
20:56 meteorfox boris-42: ok. so what level of abstraction do you see Rally working at? Just clouds, or any performance testing ?
20:56 boris-42 meteorfox: any perf testing
20:57 boris-42 meteorfox: just write random python code in context/scenarios and run them
20:57 boris-42 under load
20:57 meteorfox boris-42: hmm I see. that reminds of this project. https://github.com/aurbroszniowski/Rainfall-core
20:57 meteorfox it's java
20:57 meteorfox but it is similar idea
20:58 meteorfox boris-42: here then, use it for http perf testing: https://github.com/aurbroszniowski/Rainfall-web
20:59 psd_mo joined #openstack-rally
20:59 boris-42 meteorfox: so I didn't use it
20:59 boris-42 meteorfox: but I don't see big difference between current state of Rally and what I am propsing
21:00 boris-42 meteorfox: because if you take a look validation is the last thing that blocks stuff
21:00 boris-42 meteorfox: to use rally for HTTP testing
21:00 boris-42 meteorfox: like this https://github.com/stackforge/rally/blob/master/rally/benchmark/scenarios/requests/http_requests.py#L28
21:01 boris-42 meteorfox: so I believe that we can have a lot of stuff inside rally (sla checkers, runners, common scenarios (http/rpc/...), common context (cleanup))
21:02 boris-42 davideagnello: hm
21:02 boris-42 davideagnello: what you are returning from scenario?
21:03 boris-42 meteorfox: I mean this won't be fast decision, but we should think about this
21:04 sputnik13 joined #openstack-rally
21:21 oanufriev1 joined #openstack-rally
21:25 andreykurilin joined #openstack-rally
21:37 psd joined #openstack-rally
21:48 mwagner_lap joined #openstack-rally
21:55 meteorfox boris-42: sorry I had a meeting
22:00 boris-42 meteorfox: np
22:01 meteorfox boris-42: how's the blog post coming up?
22:02 boris-42 meteorfox: I am not writting it at the moment
22:02 meteorfox ah ok
22:12 darkhuy boris-42:this sounds very interesting...the idea of having rally be independent of openstack
22:13 darkhuy boris-42:i've actually worked on a proprietary project similar to what your talking about, but in PERL
22:13 boris-42 darkhuy: uh perl =)
22:13 darkhuy boris-42:heheh hey! perl is nice for certain things :-)
22:14 boris-42 darkhuy: I call it write only language=)
22:15 darkhuy boris-42:it was basically a benchmarking framework that allowed users to write benchmark scripts (for whatever app, scenario) they want, upload to private company RPM and it gets ran across a bunch of baremetal machines setup with a client
22:15 boris-42 darkhuy: yep quite similar to rally
22:16 darkhuy boris-42: it is probably a bit more expansive than what your thinking because we had our own file servers that hosted applications (3dmark, spec...photoshop etc) where the client machines could directly download from
22:16 prashantS joined #openstack-rally
22:16 darkhuy boris-42:but basically it would handle installation of application, run benchmark script, gather results, log to database, uninstall application, and clean up
22:17 darkhuy boris-42:i'd be really interested in working on this if that is something your going towards with rally.
22:18 boris-42 darkhuy: you mean making rally os independent?
22:18 darkhuy boris-42:it could be
22:18 boris-42 darkhuy: I know, I am refactoring it for a long amount of time
22:19 boris-42 darkhuy: to keep simplicity but make it not only for OS
22:19 darkhuy boris-42:i meant your idea of seperating rally from openstack into two repos
22:20 boris-42 darkhuy: before doing this we need to refactor all hardcoded places in rally
22:21 darkhuy boris-42:rally could become a benchmark-as a service framework
22:21 boris-42 darkhuy:  like this https://github.com/stackforge/rally/blob/master/rally/benchmark/runners/base.py#L48-L59
22:21 boris-42 darkhuy: yep exactly that
22:21 boris-42 darkhuy: with plugins for different stuff like OpenStack, Docker, ...
22:21 darkhuy boris-42:yea....feel free to assign me to some bugs fi you need help. though i'm still learning the ins and out of rally
22:22 darkhuy boris-42:is the issue with that function because it is hard-coded to reference tenants and its keys?
22:24 boris-42 darkhuy: yep
22:24 boris-42 darkhuy: so we need to extend base.Context somehow
22:25 boris-42 darkhuy: to add abbility to do such kind of mapping dynamically
22:25 boris-42 darkhuy: so users context will somehow specify the function of transforming his keys (tenants and users) to tenant and user
22:32 boris-42 darkhuy: so there are few such places
22:33 boris-42 darkhuy: in Rally code that are blocking splitting to 2 repos
22:34 darkhuy boris-42: i think i understand what your saying, but can you ELI5 to me? So users context will need to somehow identify that the key "tenant" needs to actually point to the tenant object dynamically?
22:34 darkhuy boris-42:without the need for hard-coded statements
22:35 boris-42 darkhuy: yep exactly that
22:36 boris-42 darkhuy: so function that transforms whole context to scenario context should be in base.ContextManager
22:36 boris-42 darkhuy: that can understand how to translate this stuff
22:37 boris-42 so in context dict that it get's it has already information about all context that it uses
22:37 boris-42 so we can make something like function in base.Context
22:39 boris-42 get_context_for_scenario(self, context):
22:39 boris-42 return {"user": radom.choice(context["users"]), "tenant": tenant_for_user}
22:39 boris-42
22:39 boris-42 so if context class adds new keys to context object it should override this function and return somehow to scenario part or whole information
22:40 boris-42 darkhuy: ^
22:45 darkhuy boris-42:can you explain random.choice? why random?
22:46 boris-42 darkhuy: so you are chosing random user from created to perform action
22:46 darkhuy2 joined #openstack-rally
22:46 boris-42 darkhuy: but if we do this change we will be able to have different strategies
22:49 boris-42 darkhuy: as first step I would just keep old behaviour
22:52 darkhuy2 boris-42: need to lock myself in a room for a few months to read all this rally code.... Haha... It is very overwhelming
22:53 boris-42 darkhuy2: lol I have that feeling like it's so small =)
22:53 boris-42 darkhuy2: I mean it's like my baby=) it will be always so small=)
22:56 openstackgerrit Boris Pavlovic proposed stackforge/rally: Finish support of benchmarking with existing users  https://review.openstack.org/168524
22:56 openstackgerrit Boris Pavlovic proposed stackforge/rally: Fix rally show commands when we have existing users in deployment  https://review.openstack.org/168636
22:56 openstackgerrit Boris Pavlovic proposed stackforge/rally: Delete only own resource in generic cleanup  https://review.openstack.org/174143
23:04 darkhuy boris-42:so anything specific i should start looking at?
23:05 boris-42 darkhuy: you can take this task
23:06 boris-42 darkhuy: if you are brave =)
23:10 darkhuy boris-42:the refactoring user/tenant one?
23:10 darkhuy boris-42:or you mean someting else
23:15 boris-42 darkhuy: refactoring user/tenant
23:19 darkhuy boris-42:i can try...not sure how successful i will be :-D
23:21 boris-42 darkhuy: so just 3 main files
23:21 boris-42 darkhuy: https://github.com/stackforge/rally/blob/master/rally/benchmark/context/base.py
23:21 boris-42 darkhuy: https://github.com/stackforge/rally/blob/master/rally/benchmark/runners/base.py#L48-L59
23:21 boris-42 darkhuy: and https://github.com/stackforge/rally/blob/master/rally/benchmark/engine.py#L210-L247
23:22 jaypipes joined #openstack-rally
23:25 darkhuy boris-42:cool...ill take a look this week.....there goes my weekend :-)
23:25 boris-42 darkhuy: =)
23:31 klindgren_ joined #openstack-rally
23:35 frickler_ joined #openstack-rally
23:38 aarefiev_ joined #openstack-rally
23:38 bkopilov joined #openstack-rally

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