Perl 6 - the future is here, just unevenly distributed

IRC log for #openstack-rally, 2014-08-13

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

All times shown according to UTC.

Time Nick Message
00:07 yingjun joined #openstack-rally
00:19 nkhare joined #openstack-rally
01:03 lordd_ joined #openstack-rally
01:12 oanufriev joined #openstack-rally
01:26 lordd_ joined #openstack-rally
01:32 lordd_ joined #openstack-rally
01:37 lordd_ joined #openstack-rally
01:45 oanufriev joined #openstack-rally
01:55 yaguang joined #openstack-rally
02:03 oanufriev joined #openstack-rally
02:52 shiyanch joined #openstack-rally
03:05 Poornima joined #openstack-rally
03:12 lordd_ joined #openstack-rally
03:20 vaidy joined #openstack-rally
03:30 AlexF joined #openstack-rally
04:12 vaidy boris-42, ping
04:48 Poornima joined #openstack-rally
05:13 gema_ joined #openstack-rally
05:19 ajayaa joined #openstack-rally
05:21 gema joined #openstack-rally
05:25 flwang_ joined #openstack-rally
05:30 k4n0 joined #openstack-rally
05:33 nkhare joined #openstack-rally
05:46 yingjun joined #openstack-rally
05:53 openstackgerrit Swapnil Kulkarni proposed a change to stackforge/rally: Add Rally create keypair CLI command  https://review.openstack.org/98992
05:53 coolsvap oanufriev, hi
05:54 coolsvap boris-42, k4n0 please have a look at https://review.openstack.org/#/c/113011/
05:54 k4n0 coolsvap, sure
05:54 coolsvap yingjun,  ^^
05:54 yingjun coolsvap hi
05:55 coolsvap yingjun, i updated the the change set for refactoring quotas, please have a look
05:55 yingjun coolsvap sure
05:56 AlexF joined #openstack-rally
06:21 yingjun joined #openstack-rally
06:32 openstackgerrit Swapnil Kulkarni proposed a change to stackforge/rally: Add Rally create keypair CLI command  https://review.openstack.org/98992
06:36 yingjun joined #openstack-rally
06:42 nkhare any example where we can access context variables in the benchmark
06:45 nkhare boris-42, k4n0 ^
06:49 asalkeld joined #openstack-rally
06:50 yingjun_ joined #openstack-rally
06:59 yingjun joined #openstack-rally
07:14 yingjun joined #openstack-rally
07:17 yingjun joined #openstack-rally
07:27 Alexei_987 joined #openstack-rally
07:30 yingjun joined #openstack-rally
07:36 yingjun joined #openstack-rally
07:39 yingjun joined #openstack-rally
08:06 ajayaa joined #openstack-rally
08:08 openstackgerrit Mikhail Dubov proposed a change to stackforge/rally: Add atomic action names tracking  https://review.openstack.org/111977
08:37 rediskin joined #openstack-rally
08:42 boris-42 nkhare ?
08:43 boris-42 nkhare this is the sample https://github.com/stackforge/rally/blob/master/rally/benchmark/scenarios/vm/utils.py#L56
08:44 boris-42 nkhare so everything that you put in self.context dict during creation of conetxt
08:44 boris-42 nkhare will be accessible in run_command self.context()
08:44 boris-42 vaidy hi there
08:45 vaidy boris-42, i have made some changes to my code now how would i validate if its running fine?
08:45 vaidy boris-42, sorry to ask basic questions.. just trying to get a hang of it
08:45 boris-42 vaidy there are few ways
08:46 boris-42 vaidy locally, you should install rally with pyhon setup.py install
08:46 boris-42 vaidy and try to run it against some openstack cloud
08:46 boris-42 vaidy remotely you can send patch on review
08:47 boris-42 vaidy and it will be tested by CI
08:48 vaidy ok
08:48 vaidy boris-42, i want to do it locally first to get it validated before i send it out
08:49 boris-42 vaidy ok so first of all you should run unit tests
08:49 vaidy ok
08:49 boris-42 vaidy to do this you'll need tox
08:49 boris-42 tox == 1.6.1
08:49 boris-42 then just in root directory run "tox"
08:50 boris-42 if unit tests passed you should try to run rally against some cloud
08:50 vaidy boris-42, yes i have ready about it
08:50 vaidy read*
08:50 boris-42 btw you don't need to install somewhere else rally
08:51 boris-42 you can use rally from .tox/py26
08:51 boris-42 e.g.
08:51 vaidy boris-42, ok the same place where i have modified can be tested against the openstack instance
08:51 boris-42 . .tox/py26/bin/activate
08:51 boris-42 ^ vaidy  you will need to do this, then you don't need to reinstall rally every time
08:51 vaidy boris-42, ok
08:52 vaidy boris-42, hmmm will try it with tox then
08:52 vaidy boris-42, it would be easy to debug and re-run the modified code
08:52 boris-42 vaidy +1`
08:52 boris-42 vaidy =)
08:52 vaidy boris-42, thanks
09:02 yingjun joined #openstack-rally
09:12 penguinRaider_ joined #openstack-rally
09:33 AlexF joined #openstack-rally
09:48 flwang_ joined #openstack-rally
10:07 penguinRaider_ joined #openstack-rally
10:09 amaretskiy joined #openstack-rally
10:27 rediskin about "not allowed to create an interface on external network"
10:27 rediskin https://github.com/stackforge/rally/blob/master/rally/benchmark/scenarios/nova/utils.py#L105-L107
10:27 rediskin this code is trying to add networks
10:28 rediskin which are not in configuration
10:28 rediskin and probably which should not be in configuration
10:28 rediskin this lead to "not allowed to create an interface on external network"
10:28 rediskin how it was worked before? %)
10:29 boris-42 rediskin we didn't test it in gates
10:29 rediskin boris-42: ^
10:29 rediskin T_T
10:29 boris-42 rediskin so maybe this didn't work=)
10:29 boris-42 rmk ^
10:30 penguinRaider_ joined #openstack-rally
10:42 nkhare boris-42, thanks
10:51 oanufriev joined #openstack-rally
10:52 andreykurilin_ joined #openstack-rally
10:56 tzabal joined #openstack-rally
11:07 openstackgerrit A change was merged to stackforge/rally: Refactor benchmark/context/quota.py  https://review.openstack.org/113011
11:21 scroiset joined #openstack-rally
11:30 tzabal I am trying to run a benchmark scenario (GlanceImages.list_images) that in its configuration file has the context "images", but it gives me an error. I am asking because in rally.yaml there is no scenario that actually use this context. Can someone please confirm that can run a benchmark scenario with this context defined using the latest rally snapshot?
11:32 tzabal The error seems to raise from validation, http://paste.openstack.org/show/94333/
11:34 olkonami joined #openstack-rally
11:53 flwang_ joined #openstack-rally
11:57 lordd_ Hi tzabal, could you paste your task?
11:58 lordd_ also: I've seen this task somewhere and was very curious how to use it:
11:58 lordd_ http://paste.openstack.org/show/9idRC4LeTX0TNs2BJs77/
11:58 tzabal lordd_ hi, you mean the configuration file?
11:58 lordd_ yep
11:58 tzabal lordd_ http://paste.openstack.org/show/94333/
11:58 tzabal lordd_ the last line that says benchmark configuration file
11:58 lordd_ yeah, I mean the yaml or json
11:58 tzabal lordd_ one moment
11:59 tzabal lordd_ http://paste.openstack.org/show/94352/
11:59 tzabal lordd_ it is actually the one that ships with rally under doc
12:00 tzabal lordd_ i think that the problem lies in the validate_semantic method of the images context
12:00 lordd_ I remember that there was an issue with glance
12:00 lordd_ let me check
12:00 tzabal lordd_ sure, thanks for checking that
12:01 lordd_ don't worry, I've been all day moving around html reports from the different loops
12:01 tzabal lordd_ :)
12:05 ajayaa joined #openstack-rally
12:05 lordd_ tzabal, let me see the main script for this task, as the context is completely different from other glance tasks
12:06 tzabal lordd_ you mean the benchmark scenario method list_images?
12:07 lordd_ just to see if the context is the same
12:08 tzabal lordd_ well this is the scenario method https://github.com/stackforge/rally/blob/619e77fe9f5e77bdcaa271fba3b1afcf6dcfc9d1/rally/benchmark/scenarios/glance/images.py#L53
12:08 tzabal lordd_ and this is the context images https://github.com/stackforge/rally/blob/619e77fe9f5e77bdcaa271fba3b1afcf6dcfc9d1/rally/benchmark/context/images.py
12:09 tzabal lordd_ as i am doing tests now, the validate_semantic is the one that raises the exception
12:10 nkhare joined #openstack-rally
12:12 tzabal lordd_ so it seems that this line causes the problem https://github.com/stackforge/rally/blob/619e77fe9f5e77bdcaa271fba3b1afcf6dcfc9d1/rally/benchmark/context/images.py#L109
12:13 tzabal lordd_ with glance = osclients.Clients(users[0].endpoint).glance() it seems to work
12:14 lordd_ hmm
12:14 temujin joined #openstack-rally
12:14 tzabal lordd_ endpoint is an attribute of Clients
12:15 lordd_ in that case cleanup should be different as well
12:15 tzabal lordd_ hm why?
12:17 RaginBajin joined #openstack-rally
12:18 lordd_ it is true that in some working cases, it is defined like this: osclients.Clients(endpoint)
12:18 temujin boris-42, rediskin ping
12:18 rediskin pong
12:19 lordd_ tzabal I'm puzzled why in the images.py was defined otherwise
12:19 tzabal lordd_ dunno
12:20 tzabal lordd_ the thing is that there is not a benchmark scenario in rally.yaml that actually uses this context in order to verify that it works
12:21 lordd_ tzabal don't think so. Well it looks like this might be a mighty launchpad ticket!
12:21 tzabal lordd_ yes
12:23 lordd_ tzabal These changes were made by Lingxian Kong, maybe he can shed some light on this
12:25 lordd_ In any case, if somebody has seen this context for a task with NovaServers, please ping me:
12:25 lordd_ "neutron_network": {
12:25 lordd_ "network_cidr": "10.%s.0.0/16",
12:26 tzabal lordd_ i haven't seen it
12:26 rook-out noe
12:26 rook-out nope*
12:26 lordd_ thx
12:27 tzabal lordd_ this task actually runs?
12:27 rediskin k4n0: sup
12:28 rediskin k4n0: could you pls take a look https://review.openstack.org/#/c/110007/6/rally/benchmark/runners/rps.py
12:28 k4n0 rediskin, nm, busy with some internal work
12:28 k4n0 rediskin, sure
12:28 rediskin k4n0: ok, sorry
12:28 tzabal lordd_ as far as i know, in order to run it has to be an existing context, and all the existing contexts are located in rally/benchmark/context
12:28 k4n0 rediskin, I agree with your comment though,
12:28 tzabal lordd_ and i cannot find it there
12:31 lordd_ tzabal then somebody got creative! The funny thing is it was in a tutorial for Rally
12:32 tzabal lordd_ oh
12:32 boris-42 joined #openstack-rally
12:42 boris-42 k4n0 please could you re-review RPS patch
12:42 k4n0 boris-42, sure
12:42 boris-42 k4n0 as rediskin is quite busy we decided to refactor unit tests in next patch
12:42 k4n0 boris-42, ohh, ok
12:42 boris-42 k4n0 Oleg is going to do that
12:43 boris-42 nkhare hi
12:43 boris-42 nkhare could you please address this issues https://review.openstack.org/#/c/107595/
12:43 k4n0 boris-42, lgtm
12:44 boris-42 nkhare for example now we have inside rally functionality to reduce amount of points
12:44 boris-42 nkhare so what you need is just to replot results
12:45 boris-42 k4n0 you forgot +1 to workflow
12:48 nkhare boris-42, I might not have all result. I'll check and replot
12:49 boris-42 nkhare if so probably it makes sense to remove graphs =(
12:49 boris-42 nkhare cause they are killing browsers..
12:49 nkhare boris-42, yes.
12:49 boris-42 nkhare or probably rediskin can make some trick
12:49 boris-42 rediskin to reduce amount of points on existing graphs
12:49 nkhare I'll confim tomorrow
12:49 nkhare boris-42, ^
12:50 rediskin on existing graphs?
12:50 rediskin parse and regenerate html/json?
12:50 rediskin T_T
12:51 AlexF joined #openstack-rally
12:51 rediskin http://i720.photobucket.com/albums/ww205/Solen_Inthuul/Dkiw.jpg
12:53 boris-42 rediskin ^_^
12:54 nkhare :)
13:02 aswadr joined #openstack-rally
13:04 boris-42 aswadr hi
13:04 aswadr Hi boris-42
13:05 boris-42 aswadr how are you?
13:05 aswadr boris-42: I am fine, thanks
13:05 aswadr boris-42: how are you ?
13:05 boris-42 aswadr heh good=)
13:06 boris-42 aswadr a bit tired=)
13:07 aswadr boris-42: lot of work...i guess for the summit. How is the preps going
13:08 boris-42 aswadr nope actually it's more about trying to get rally a official part of openstack..
13:11 penguinRaider_ tzabal, I think this is perhaps because this line defines users of validate_semantic arg as an endpoint object https://github.com/stackforge/rally/blob/master/rally/benchmark/engine.py#L151
13:12 boris-42 aswadr are you working now on any task?
13:17 tzabal penguinRaider_ ok, but what do you mean by that?
13:20 tzabal penguinRaider_ the users parameter inside the validate_semantic method it seems to be a Clients object
13:20 ajayaa joined #openstack-rally
13:27 penguinRaider_ tzabal, yeah you are right its a Clients objects what I meant was if you look at the flow of the program the users arg of validate_semantic is Clients object because of that line IMO
13:28 boris-42 penguinRaider http://logs.openstack.org/19/113119/4/check/rally-coverage/a678358/cover/rally_benchmark_scenarios_requests_http_requests.html
13:28 boris-42 penguinRaider could you cover red line with tests ^
13:28 tzabal penguinRaider_ ah yes, ok with that.
13:28 boris-42 penguinRaider as well could you at least call one time in gates
13:29 boris-42 penguinRaider https://github.com/stackforge/rally/blob/master/rally-scenarios/rally.yaml put somewhere here one of calls to it
13:29 penguinRaider_ boris-42, yeah okay
13:51 boris-42 aswadr wanna some small task?
13:52 aswadr boris-42: sure
13:52 aswadr boris-42: but i am caught up with a task in office right now
13:52 boris-42 aswadr it's not urgent
13:52 boris-42 aswadr just nice to have
13:53 aswadr boris-42: can I ping you once I reach home and pick it up?
13:53 aswadr sure Boris
13:53 boris-42 aswadr heh I can say now for example
13:53 boris-42 aswadr https://github.com/stackforge/rally/blob/master/rally/benchmark/scenarios/keystone/utils.py#L45-L48
13:53 boris-42 aswadr you need to use instead of @scenario_base.atomic_action_timer("keystone.delete_resource")
13:53 openstackgerrit A change was merged to stackforge/rally: Fix partial results sending in rps runner  https://review.openstack.org/110007
13:54 boris-42 aswadr https://github.com/stackforge/rally/blob/master/rally/benchmark/scenarios/base.py#L236
13:54 boris-42 aswadr with AtomicAction(self, name_of_resource):
13:54 flwang_ joined #openstack-rally
13:54 boris-42 aswadr so we will get in keystone benchmark not "deleted_resource"
13:54 boris-42 aswadr but exactly the name of resource that was deletede
13:55 aswadr boris-42: ok
13:57 boris-42 coolsvap hey there
13:57 coolsvap boris-42, hi
13:58 boris-42 coolsvap could you please finish work on this https://review.openstack.org/#/c/112973/ ?
13:59 boris-42 coolsvap it shouldn't take too much time
13:59 boris-42 coolsvap there are just to places that shouldn't be changed
13:59 coolsvap boris-42,  sure will do
13:59 boris-42 coolsvap thank you
14:03 olkonami left #openstack-rally
14:03 ajayaa joined #openstack-rally
14:13 openstackgerrit Swapnil Kulkarni proposed a change to stackforge/rally: Update the scenario_base to base in benchmarks  https://review.openstack.org/112973
14:13 coolsvap boris-42, ^^
14:15 boris-42 coolsvap thank you
14:15 boris-42 coolsvap let me review first your patch related to @validation stuff
14:15 coolsvap boris-42, yup
14:16 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Fix semantic validation of context "images"  https://review.openstack.org/113904
14:17 openstackgerrit joined #openstack-rally
14:29 rook So I can set a concurrency, but can I define a # of VMs i would like to see created?
14:29 rook so, guests launch 10 guests at a time, but go to 1000 guests?
14:32 boris-42 rook hm elaborate please..
14:33 boris-42 rook concurrency is only the parameter of one of load generators..
14:34 rook boris-42 I can define the concurrency for a scenario, but I want to define a # of widgets... where # of widgets could be 1000, so with a concurrency of 10, the workload would run 100 times, to reach 1000 widgets.
14:35 boris-42 rook hmmm times & concurrency are not related arguments
14:35 boris-42 rook in constant runner
14:35 boris-42 rook times -- total amount of iterations, concurrency - amount of iterations that load generator should keep in the every moment
14:35 rook Right every time I reach 10 guests, and the workload is complete, it tears down and decrements from the # of times.
14:36 boris-42 rook so basically you would like just to create VMs and not delete them?
14:37 rook I would like to build up to a specific number of guests/widgets, instead of stating I want 100 guests launched.
14:37 rook I want a specific concurrency
14:37 rook and a end goal of # of widgets/guests
14:38 rook From what I am hearing, this is not currently possible
14:38 boris-42 rook so then you need next
14:38 boris-42 rook everything is possible lol
14:38 rook HAHA
14:39 boris-42 rook you should use benchmark scenario that just boot VM (and don't delete them)
14:39 rook mightily optimistic today!
14:39 boris-42 rook https://github.com/stackforge/rally/blob/master/rally/benchmark/scenarios/nova/servers.py#L145
14:39 boris-42 rook then you should use constant runner
14:39 boris-42 rook specify concurrency 10 and times 1000
14:39 boris-42 rook you will create 10 VMs per any moment, and at the end you'll get 1000 running VMs
14:39 boris-42 rook is that what you would like to do?
14:40 rook roger - simular yes.
14:40 boris-42 rook but this VMs will be cleaned up
14:40 boris-42 rook after load generator finish work
14:40 boris-42 rook in cleanup context
14:40 boris-42 these VMs*
14:41 rook makes sense, let me play with this idea.
14:41 rook Thanks boris
14:42 boris-42 rook btw you'll definitely need to setup quotas
14:42 rook oh yeah
14:42 boris-42 rook it's as well possible
14:42 boris-42 rook lemme point you
14:42 rook i set up the neutron quotas through my yaml
14:42 boris-42 rook ah so you know about quotas context?
14:42 boris-42 rook so here are nova quotas https://github.com/stackforge/rally/blob/master/rally/benchmark/context/quotas/nova_quotas.py#L27-L75
14:43 rook Yup
14:43 rook boris-42++
14:43 rook thanks
14:43 boris-42 rook ok great
14:46 andreykurilin_ joined #openstack-rally
14:47 tnurlygayanov joined #openstack-rally
14:48 ajayaa joined #openstack-rally
15:01 rediskin left #openstack-rally
15:15 coolsvap boris-42, andreykurilin_ https://review.openstack.org/#/c/112973/ please check once
15:15 boris-42 coolsvap you definitely like big patches, that changes a lot =)
15:16 coolsvap boris-42, not really I actually get caught up with some fixes while fixing some fix
15:16 coolsvap boris-42, i will try to split up in future
15:20 openstackgerrit A change was merged to stackforge/rally: Added Sahara Clusters scenario  https://review.openstack.org/109557
15:26 boris-42 coolsvap could you update your validation patch?
15:26 boris-42 coolsvap tzabal refactored it a bit
15:26 boris-42 coolsvap so now it's quite out of date
15:27 coolsvap boris-42, I think i will need to update https://review.openstack.org/#/c/112973/ again :-/
15:27 coolsvap boris-42, i will update the validation patch
15:27 boris-42 coolsvap why do you need to update it one more time?
15:28 coolsvap i think https://review.openstack.org/109557 merge can cause conflicts jut thinking
15:31 boris-42 coolsvap ok
15:31 boris-42 coolsvap then rebase it
15:31 boris-42 coolsvap and let's merge it
15:31 coolsvap boris-42, doing it
15:55 flwang_ joined #openstack-rally
16:03 openstackgerrit Swapnil Kulkarni proposed a change to stackforge/rally: Update the scenario_base to base in benchmarks  https://review.openstack.org/112973
16:04 boris-42 ^ https://review.openstack.org/#/c/112973/4/tests/benchmark/runners/test_base.py coolsvap ohhhh
16:04 coolsvap boris-42, checking what happened with this
16:04 coolsvap 1 min
16:07 andreykurilin_ joined #openstack-rally
16:13 openstackgerrit Swapnil Kulkarni proposed a change to stackforge/rally: Update the scenario_base to base in benchmarks  https://review.openstack.org/112973
16:14 coolsvap boris-42, i had skipped it
16:14 boris-42 coolsvap yep that looks better=)
16:14 coolsvap but somehow managed to get it in which caused issue
16:14 boris-42 coolsvap let's wait gates and merge it
16:14 coolsvap boris-42, yup
16:14 coolsvap boris-42, i will update the validation patch in sometime, need to cook dinner will be afk for some time
16:15 boris-42 coolsvap bon appétit =)
16:32 Gamekiller77 joined #openstack-rally
16:48 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Fix semantic validation of context "images"  https://review.openstack.org/113904
16:59 Gamekiller77 joined #openstack-rally
17:03 AlexF joined #openstack-rally
17:09 andreykurilin_ joined #openstack-rally
17:20 rmk boris-42: That isn't even code I am adding.
17:20 rmk I saw that in the review.  If there's an issue with that code it needs to be a separate bug.
17:20 boris-42 rmk ya your just cleanping
17:20 rmk Reviewing code I didn't add/change, except to make it conditional based on other logic, seems strange.
17:21 boris-42 rmk I think we should accept your clenap
17:21 rmk There are a few proper nits I need to sort
17:21 boris-42 rmk heh I mean it's not about your patch=)
17:21 rmk And if there's an issue with that floating ip section then that's a separate bug imo
17:21 boris-42 rmk we have just issues=)
17:21 rmk boris-42: Yeah there was a review on my patch for it
17:21 boris-42 rmk sorry about that=)
17:21 rmk It's fine I'm just saying it's a separate bug, we should fix it
17:21 rmk No big deal
17:21 boris-42 rmk +1=)
17:21 boris-42 rmk sorry about that=)
17:21 rmk No need to apologize
17:22 boris-42 we are just trying to review as best as we can=)
17:22 rmk Im glad that part of the code is getting attention, it was my first blocker when I started using rally
17:24 boris-42 rmk so can you just update that patch
17:24 boris-42 rmk that fixes one crap
17:24 rmk yep
17:24 boris-42 rmk and we will continute on fixing another crap=)
17:24 rmk will do it shortly
17:24 boris-42 rmk or you can if you would like=)
17:24 boris-42 rmk ok nice
17:24 rook boris-42 wasn't there a scenario for iperf?
17:25 boris-42 rook not yet
17:25 rook boris-42 i just wrote one for netperf, but i haven't pushed it upstream yet
17:25 rook ok
17:25 boris-42 rook tzbal is working hard on that
17:25 boris-42 tzabal *
17:25 rook I just adapted VMTask.py
17:25 boris-42 rook you can do a separated benchmark
17:26 boris-42 rook that will start N VMs
17:26 boris-42 rook and run IPERF
17:26 boris-42 rook between them
17:26 boris-42 rook based on VMTask.py
17:26 boris-42 rook what do you think?
17:26 rook Right, I have that now - but i put it in VMTask.py
17:26 boris-42 rook it's okay to keep benchmark in VMTask.py
17:27 boris-42 rook just call it something like def iperf()
17:27 boris-42 rook we need to get some your contribution to rally!
17:27 boris-42 rook sorry
17:27 boris-42 rook I meant rmk!=)
17:27 boris-42 rook but Iperf stuff will be as well very nice to get
17:28 boris-42 rook btw do you know about this https://rally.readthedocs.org/en/latest/concepts.html#scenarios-plugins
17:28 rook IMHO the code is ugly - it is replicating VMTask - since the code would worth with both iperf and NetPerf I just created : def boot_two_runcommand_delete
17:28 boris-42 rook oh
17:29 boris-42 rook that can be refactored=)
17:30 rook would work*
17:30 boris-42 rook thank you
17:31 tzabal rook do you have the netperf under review or in your local environment?
17:31 rook local
17:32 tzabal rook have you seen this one about blogbench? https://rally.readthedocs.org/en/latest/concepts.html#scenarios-plugins
17:32 tzabal rook sorry wrong link :P, https://review.openstack.org/#/c/97030/
17:32 rook i haven't, i saw messages about it, haven't tkaen the time to look
17:32 tzabal rook ok
17:32 tzabal rook i read about refactoring, i am not sure if you were talking about VMTasks
17:33 tzabal rook and i think that we need to separate the boot-check network-access vm-execute command- in a generic method in vmtasks
17:33 tzabal rook in order to be used by many places
17:35 tzabal rook anyway, if you have the intention to upload netperf, it would be nice to spea and unify the work
17:35 rook well... the work i did for netperf was quick and dirty
17:35 tzabal rook ok then
17:36 rook https://gist.github.com/anonymous/d4cca6f63933997f5cd1
17:37 tzabal rook i see
17:37 rook https://gist.github.com/anonymous/6f7d887c5eaf9597a6b5
17:37 rook that was the script it would point to
17:37 rook it could easily be changed to iperf.
17:37 tzabal rook great
17:38 tzabal rook the same logic was when i was working in earlier patch sets for blogbench
17:39 tzabal rook so as soon as we decide on a final architecture, netperf with the current state that you have it, it wont need much to be ported
17:39 coolsvap boris-42, andreykurilin_ two workflow+1 to the same patch ;)
17:40 rook tzabal well - I don't think we need to do very much for such a small workload
17:40 boris-42 coolsvap double success!=)
17:41 rook tzabal if we want to introduce like YCSB and run it against trove or something... that might constitute a look.
17:41 rook tzabal: since netperf is pretty simple, I don't think we need to do much.... That is why I hesitate to push it upstream.
17:42 tzabal rook sure ok
17:42 rook if you tzabal and boris-42 think it would add value, ill do it... no problem there =)
17:42 tzabal rook with YCSB you mean Yahoo Cloud Serving Benchmark or something else? :)
17:43 rook Yeah, Yahoo Cloud Service Benchmark
17:43 rook i use that for NOSQL db testing
17:43 tzabal rook nice
17:47 andreykurilin_ guys. I'm going on vacation until august 23. I will be available via e-mail akurilin@mirantis.com
17:49 rook boris-42: would you like to see the netperf code pushed, or should I hold out until tzabal's fixes are in?
17:49 rook it really isn't a "netperf" scenario
17:50 rook it is just a scenario where you have two guests, one guest running a workload against a service running on the other.
17:51 boris-42 rook I would like to see this in code
17:51 boris-42 rook if you are asking me
17:51 rook i posted the code
17:51 boris-42 rook cause it is one thing when user can just run yaml
17:51 boris-42 rook and another thing when user need to write own benchmark
17:52 boris-42 rook even if it is simple
17:52 rook i don't consider netperf a "benchmark" but sure =)
17:52 boris-42 rook so you'll need to use gerrit and OpenStack CI stuff
17:52 boris-42 rookhttps://wiki.openstack.org/wiki/Rally/Develop#How_to_contribute
17:52 boris-42 rook https://wiki.openstack.org/wiki/Rally/Develop#How_to_contribute *
17:52 rook oh, i have already contributed...
17:53 rook i know the pain... process... ;)
17:53 rook you were out vacationing.
17:54 rook tzabal: does y our work assume a single guest?
17:54 rook looking through the code, it seems to be that way ^
17:56 boris-42 rook hehe
17:56 boris-42 rook ya first time is hard lol
17:56 flwang_ joined #openstack-rally
17:56 boris-42 rook to contirubate
17:59 tzabal rook at this point yes
17:59 rook ah, bummer
17:59 rook ok
18:00 tzabal rook the main part is the method benchmark inside VMTasks
18:00 rook Right, that is what I looked throguh
18:00 rook through, to determine that tzabal
18:01 rook tzabal: this is good stuff
18:01 tzabal rook the thing is that benchmark method is essentially what boot_runcommand_delete needs to perform, what blogbench needs to perform, and every benchmark that requires only one vm
18:02 tzabal rook so before this thing get merged, for sure it will be refactored to be available to any method that wants this functionality
18:02 tzabal rook so in case of netperf, that needs 2 guests, maybe we need to add another benchmark method
18:02 tzabal rook or refactor in some way benchmark, to be able to boot N servers
18:03 tzabal rook but maybe this is a long shoot for now, at least for me
18:03 rook tzabal: there just isn't a good way of doing that currently IMHO
18:03 tzabal rook probably yes
18:05 rook hm
18:05 rook tzabal: thinking about it.
18:05 rook tzabal: we could have a method boot_runcommand (for the infra guest), then boot_runcommand that will be the last guest, which will run the workload.
18:06 rook so, it isn't boot_runcommand_delete
18:06 rook let me see if I can clean things up a bit, with my netperf example
18:07 tzabal rook ok
18:09 rook tzabal: we could have a prepare_benchmark function, that will stand up the guests and start services of wahtever we are going to benchmark (netserver, iperf -s or whatever), then a run_benchmark that will execute and return the results, then cleanup
18:10 openstackgerrit A change was merged to stackforge/rally: Update the scenario_base to base in benchmarks  https://review.openstack.org/112973
18:11 tzabal rook about the first step, the prepare_benchmark function, if i understood correctly your thought, it will be where we install and have everything ready for the benchmarking to be done
18:12 rook no
18:12 rook tzabal: my assumption is netperf is already installed on the guest
18:12 tzabal rook ok thats good
18:12 rook tzabal: with the netperf example, prepare_benchmark would be the function to start the netserver in the netperf example
18:13 tzabal rook right, it is needed for this kind of benchmark
18:13 rook it is needed for a lot of benchmarks
18:13 tzabal rook aha :)
18:13 rook distributed benchmarks**
18:13 rook tzabal single machine workloads aren't interesting ;P
18:13 tzabal rook it is true
18:14 tzabal rook look, let me upload some changes and we can refactor then the whole thing with having in mind the step of starting the services
18:18 rook tzabal: no problem
18:19 dpaterson joined #openstack-rally
18:31 Gamekiller77 joined #openstack-rally
18:49 openstackgerrit joined #openstack-rally
19:08 rmk Actually two of the comments are for changes to my patch which my patch didn't introduce
19:08 rmk I think there's way too much emphasis on coverage being 100% even when it doesn't make sense
19:24 oanufriev joined #openstack-rally
19:25 openstackgerrit Rafi Khardalian proposed a change to stackforge/rally: Do not assume networks should be assigned  https://review.openstack.org/113096
19:26 rook_ joined #openstack-rally
19:36 boris-42 rmk +1
19:37 boris-42 rmk will wait for Jenkins and review your pathc
19:37 rmk cool thanks
19:40 coolsvap boris-42, I did the refactoring for validator again
19:41 coolsvap got some idea about tests, do you have the tests?
19:41 coolsvap else I can write them tomorrow
19:43 RaginBaj_ joined #openstack-rally
19:44 gema_ joined #openstack-rally
19:53 boris-42 coolsvap hm
19:53 boris-42 coolsvap but there are already tests no?
19:57 flwang_ joined #openstack-rally
20:08 gema joined #openstack-rally
20:16 gema_ joined #openstack-rally
20:30 boris-42 rmk around?
20:31 boris-42 rmk why not just passing 2 benchmarks here https://review.openstack.org/#/c/113096/5/rally-scenarios/rally.yaml
20:31 boris-42 rmk ?
20:31 gema joined #openstack-rally
20:34 AlexF joined #openstack-rally
20:34 rmk you mean true?
20:35 boris-42 rmk ya I mean just have 2 benchmarks
20:35 boris-42 rmk lemme just paste
20:35 rmk Ah I see
20:35 rmk one with and one without
20:35 boris-42 rmk ya
20:35 rmk ok
20:35 boris-42 rmk as a array
20:35 boris-42 rmk then we have better functional coveregate
20:47 gema_ joined #openstack-rally
20:54 gema joined #openstack-rally
20:55 asalkeld joined #openstack-rally
21:04 gema_ joined #openstack-rally
21:10 Gamekiller77 joined #openstack-rally
21:22 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Add the context "benchmark_image"  https://review.openstack.org/104564
21:44 kevtsang joined #openstack-rally
21:52 jaypipes joined #openstack-rally
21:53 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Add the base method for the benchmarking of VMs  https://review.openstack.org/98172
21:55 lordd_ joined #openstack-rally
21:56 boris-42 kevtsang hi there
21:58 flwang_ joined #openstack-rally
22:14 RainbowBastion joined #openstack-rally
22:15 RainbowBastion boris-42 Hi, Gamekiller77 and I wanted to ask you about your workaround to the LDAP issue
22:15 boris-42 RainbowBastion so two patches are already merged to that
22:15 boris-42 RainbowBastion there are 2 in work progress
22:16 boris-42 RainbowBastion do you have some time to contribute?
22:16 RainbowBastion Sure
22:16 boris-42 RainbowBastion you can help and go from other side
22:17 boris-42 RainbowBastion so we won't have intersection and will get job done faster
22:17 RainbowBastion boris-42 Let me know where to start
22:17 boris-42 RainbowBastion just a few questions
22:17 boris-42 RainbowBastion did you contribute anything to any of Openstack stack forge porjects?
22:18 RainbowBastion boris-42 I've been documenting the unit tests for Rally and for Tempest for the past few weeks, it's stored on an internal Cisco site at the moment
22:19 boris-42 RainbowBastion it will be nice to share this=)
22:19 boris-42 RainbowBastion https://github.com/stackforge/rally/tree/master/doc/source
22:20 boris-42 RainbowBastion so it will be nice to organize auto generation
22:20 boris-42 of all Scenarios
22:20 boris-42 e.g. imporve DOC string
22:20 boris-42 and organize some better strucktre
22:20 RainbowBastion boris-42 I'll check with Gamekiller77, since I'm an intern under his direction :P
22:21 boris-42 RainbowBastion okay but back to LDAP
22:21 RainbowBastion boris-42 yes
22:22 jaypipes joined #openstack-rally
22:24 RainbowBastion boris-42 Just talked with Gamekiller77, I'll post the information when I'm closer to finishing it, I have about 30-35 files full of tests to document
22:24 RainbowBastion that are left
22:24 boris-42 RainbowBastion huge amount of work
22:25 boris-42 RainbowBastion so how much time you can spend?
22:25 boris-42 RainbowBastion on LDAP?
22:27 RainbowBastion boris-42 Well, Gamekiller77 and I can't run any Rally tests until we get this issue resolved, so this is priority #1.  I'll spend all my work time on it as long as it gets fixed.
22:28 boris-42 RainbowBastion okay I'll give you my task (that is a bit simpler)
22:28 boris-42 RainbowBastion and start working on hard one)
22:29 boris-42 RainbowBastion so you are quite familiar with code for now?
22:29 RainbowBastion boris-42 yeah
22:29 boris-42 RainbowBastion that's great =)
22:29 RainbowBastion boris-42 if I have any questions, I'll ask
22:29 boris-42 RainbowBastion sure
22:30 boris-42 RainbowBastion so first task is next
22:30 boris-42 RainbowBastion https://github.com/stackforge/rally/blob/master/rally/benchmark/runners/base.py#L221-L224
22:30 boris-42 RainbowBastion actually https://github.com/stackforge/rally/blob/master/rally/benchmark/runners/base.py#L201-L221
22:31 boris-42 RainbowBastion we are crating context inside scenario runner, and this should be done in benchmark engine
22:31 boris-42 RainbowBastion https://github.com/stackforge/rally/blob/master/rally/benchmark/engine.py#L193
22:31 boris-42 RainbowBastion so we should create context and pass call runner with context_object
22:32 boris-42 RainbowBastion not create it inside runner
22:34 RainbowBastion boris-42 So the benchmark engine is responsible for creating a context file, and then the context file is passed to base.py to run the configuration?
22:35 boris-42 RainbowBastion it's not file
22:35 boris-42 RainbowBastion it's shared object
22:35 boris-42 RainbowBastion lemme try to explain
22:35 boris-42 RainbowBastion we are creating some kind of dict() and it goes through all context.setup()
22:36 boris-42 RainbowBastion and it is changed in every context
22:36 boris-42 RainbowBastion after that it's passed to scenario where you can consume info from it
22:36 boris-42 RainbowBastion at this point runner is creating context object, and more proper arch is to do this job in benchmark.engine
22:37 boris-42 RainbowBastion cause benchamrk.engine is glue between scenarios, runners, validation of arguments, and binding with deployment
22:37 boris-42 RainbowBastion so runner should just create a load, scenario should just run some actions from users, context should just create env
22:37 boris-42 RainbowBastion and benchmark.engine should coordinate this stuff
22:38 RainbowBastion boris-42 So benchmark engine is where the shared objects are stored, then passed around to scenario, runners, etc. in order to modify their contents?
22:38 boris-42 RainbowBastion it's not stored at least yet=)
22:39 boris-42 RainbowBastion I mean benchamrk.engine should create using context benchmark enviroment
22:39 boris-42 RainbowBastion and pass this env directly to runner
22:39 boris-42 RainbowBastion and not create env in runner
22:40 boris-42 so actually all this yellow
22:40 boris-42 more or less should be done in benchmark engine run
22:43 RainbowBastion So the steps are as follows: benchmark.engine creates the context environment through calling context.setup() (inside benchmark.engine.run()), then passes it to the runner.  Then the runner will handle the user creation to bypass the LDAP issue?
22:46 RainbowBastion boris-42 I assume this task right now is refactoring, actually, not sure how it ties into LDAP though
22:47 boris-42 RainbowBastion as I said you are doing from one side
22:47 boris-42 me from another
22:47 RainbowBastion boris-42 Fair enough
22:47 boris-42 RainbowBastion I am going to work on this https://github.com/stackforge/rally/blob/master/rally/benchmark/engine.py#L205-L217
22:47 boris-42 RainbowBastion take a look at noe
22:48 RainbowBastion boris-42 So context.setup() will generate a shared object in a JSON format?
22:48 boris-42 RainbowBastion and when you move code to engine.run() and I fix this passing of clients
22:48 boris-42 RainbowBastion nope it's just python dict
22:48 boris-42 RainbowBastion it can be json.dumps(this_dict)
22:48 boris-42 =)
22:48 boris-42 RainbowBastion so when I finish my part and you finish your
22:49 boris-42 RainbowBastion we will need to refactor last point
22:49 boris-42 RainbowBastion https://github.com/stackforge/rally/blob/master/rally/benchmark/runners/base.py#L205-L208
22:49 boris-42 RainbowBastion remove default users context, that actually creates user
22:49 boris-42 RainbowBastion and have 2 different behavior
22:50 boris-42 RainbowBastion e.g. if admin endpoint is passed, and no user via deployment => add this context if it is not added
22:50 RainbowBastion boris-42 I see, so that should be separate to make it more clear about what each function does
22:50 boris-42 RainbowBastion ya you'll refactoring from one side
22:50 boris-42 RainbowBastion me from another
22:51 boris-42 RainbowBastion and then just somebody probably you will remove this scenario_context.setdefault("users", {}_
22:51 boris-42 RainbowBastion as a result when you are creating deployment you'll be able to provide users that already exist
22:51 boris-42 RainbowBastion and use them for bnehcmarking
22:54 RainbowBastion boris-42 great
22:54 boris-42 RainbowBastion heh we are refactoing rally for more then half or year
22:55 boris-42 RainbowBastion to enable LDAP lol
22:55 RainbowBastion boris-42 Sounds like it's on its last few steps to get it working though :)
22:55 boris-42 RainbowBastion yep=)
22:56 boris-42 RainbowBastion but it was really huge work =)
22:56 boris-42 RainbowBastion e.g. we invented benchamrk.context to do this
22:56 boris-42 =)
22:58 RainbowBastion boris-42 Alright, I'll get started, would it be easier to contact you on here if I have questions or should I just send them to your email?
22:58 boris-42 RainbowBastion so I am using znc
22:58 boris-42 RainbowBastion so I am always online*
22:58 boris-42 RainbowBastion and it is better to ping me here and say, cause this channel is logged and others Rally contributors will be able to track our discussions
23:00 Gamekiller77 joined #openstack-rally
23:01 RainbowBastion boris-42 will do!
23:10 boris-42 RainbowBastion great
23:10 boris-42 RainbowBastion btw to publish your patch
23:10 boris-42 RainbowBastion use this https://wiki.openstack.org/wiki/Rally/Develop#How_to_contribute
23:23 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Add the benchmark 'Blogbench' for the Virtual Machines  https://review.openstack.org/97030
23:25 openstackgerrit OpenStack Proposal Bot proposed a change to stackforge/rally: Updated from global requirements  https://review.openstack.org/114078
23:36 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Add the benchmark 'Blogbench' for the Virtual Machines  https://review.openstack.org/97030
23:38 lordd_ joined #openstack-rally
23:56 openstackgerrit Oleh Anufriiev proposed a change to stackforge/rally: refactoring rps runnet unit tests  https://review.openstack.org/114083
23:59 openstackgerrit Oleh Anufriiev proposed a change to stackforge/rally: refactoring rps runner unit tests  https://review.openstack.org/114083
23:59 flwang_ joined #openstack-rally

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