Perl 6 - the future is here, just unevenly distributed

IRC log for #openstack-rally, 2015-03-22

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

All times shown according to UTC.

Time Nick Message
00:19 openstackgerrit Merged stackforge/rally: Updated from global requirements  https://review.openstack.org/166486
00:21 panbalag joined #openstack-rally
00:51 psd_ joined #openstack-rally
01:28 PerfBeing joined #openstack-rally
03:25 psd joined #openstack-rally
04:30 PerfBeing joined #openstack-rally
04:30 dpaterson joined #openstack-rally
04:49 dspano joined #openstack-rally
05:16 PerfBeing joined #openstack-rally
05:17 tfreger joined #openstack-rally
05:28 PerfBeing joined #openstack-rally
05:41 psd joined #openstack-rally
06:33 nmagnezi joined #openstack-rally
07:25 yfried joined #openstack-rally
07:26 yfried boris-42: ping
07:45 nmagnezi joined #openstack-rally
08:17 openstackgerrit Nir Magnezi proposed stackforge/rally: Adds multi network and IPv6 support for scenarios  https://review.openstack.org/154885
08:34 exploreshaifali joined #openstack-rally
08:53 andreykurilin_ joined #openstack-rally
09:20 exploreshaifali joined #openstack-rally
09:24 openstackgerrit Nir Magnezi proposed stackforge/rally: (WIP) Adds IPv6 Scenario to vmtasks  https://review.openstack.org/166591
11:11 boris-42 yfried: hihi
11:33 yfried boris-42: ?
11:36 boris-42 yfried: you ping me today?)
11:46 exploreshaifali joined #openstack-rally
11:56 yfried boris-42: right
11:56 yfried boris-42: some more questions
11:56 yfried if you are free
11:58 yfried boris-42: I asked you last week about the possibility of upgrading a live rally server but didn't get a answer
11:59 yfried boris-42: is it reasonable to upgrade the server every 1-2 weeks to work close to master branch?
12:00 boris-42 yfried: actually I answered
12:01 boris-42 yfried: seems like you missed my reply
12:02 boris-42 yfried: so for now we are NOT doing changes in DB
12:03 boris-42 yfried: we are trying to not remove anything instantly (so we have deprecation mechanism)
12:03 boris-42 yfried: so now I will work on DB migrations and testing of that stuff
12:06 boris-42 yfried: and about backuping rally DB
12:06 boris-42 yfried: I don't see any big issues with that..
12:06 boris-42 it's just standard practice
12:13 exploreshaifali boris-42, hello
12:14 exploreshaifali boris-42, for few days I was busy in college and other stuff, could not work for "generating trends"
12:15 exploreshaifali till april it would be same hectic schedule for me
12:15 exploreshaifali but I will submit my current patch for rally bug asap, after the modifications suggested!
12:23 boris-42 exploreshaifali: ok=)
12:23 exploreshaifali :)
12:31 yfried boris-42: I'm not sure how your reply answers my question. I wonder if rally has a "rally-update" command, or if you have tested the update process
12:32 yfried boris-42: assuming I have a live server, and I want to update it, do I need to DL the new code manually and re-run install-script, and if so - what will happen to my old DB?
12:36 boris-42 yfried: so we don't have update script yet
12:36 boris-42 yfried: if you run install_rally.sh it will purge db
12:37 boris-42 yfried: if you run python setup.py install with fresh code (it will update rally)
12:37 boris-42 or just use pip install rally
12:37 boris-42 with specific version
12:38 yfried boris-42: and how do I keep the DB intact?
12:39 boris-42 yfried: there is no changes in it
12:39 boris-42 yfried: at least between 0.0.1 and 0.0.2
12:39 boris-42 yfried: and I think we will have 0.0.3 as well without changes
12:39 yfried boris-42: " if you run install_rally.sh it will purge db"
12:39 boris-42 yfried: don't run install_rally.sh
12:39 boris-42 yfried: run "pip install " or "python setup.py install"
12:40 yfried boris-42: so pip install -U rally should pull latest version without touching the existing DB?
12:41 boris-42 yfried: it won't touch
12:41 boris-42 yfried: db at all
12:41 boris-42 yfried: there is way to do the stuff with install_rally.sh
12:42 yfried boris-42: what about scalability?
12:42 boris-42 like you can change to some other place DB in /etc/rally/rally.sh
12:42 boris-42 yfried: scalability?
12:42 yfried boris-42: how will a single rally server function running against multiple deployments at once?
12:45 boris-42 yfried: it depends on your CPU
12:45 boris-42 yfried: so the total load that you can generate is always constant
12:45 boris-42 yfried: and depends on runners that we have
12:46 boris-42 yfried: as you saw we run about 5k tasks and everything still works=)
12:47 yfried boris-42: 5k tasks in parallel?
12:47 boris-42 yfried: at all
12:47 boris-42 yfried: are you going to run 5k tasks in parallel?
12:47 boris-42 yfried: where are you going to find so much hardware?)
12:48 yfried boris-42: no, but we have ~200 jobs running in parallel and we wonder how it will work with a single rally server\
12:48 boris-42 yfried: not good
12:49 boris-42 yfried: you won't be able to generate big load in every job
12:49 boris-42 yfried: I am not sure that you can run more then 2-3k iterations in parallel
12:52 boris-42 yfried: there is a huge portion of work that we should do in Rally to scale stuff
12:52 boris-42 yfried: one of them is changes in DB (I am going to publish spec on monday)
12:52 boris-42 yfried: another is testing rally with rally
12:52 boris-42 so we will be able to get real numbers regarding to perfromance/scalability of rally
12:55 yfried boris-42: ok.
12:55 yfried boris-42: about new users (not devs) jumping into rally
12:56 boris-42 yfried: ?
12:56 yfried boris-42: the CLI "info" tool is pretty cool about giving info re specific task
12:56 yfried boris-42: but what about pulling everything into readthedocs?
12:57 yfried boris-42: so you could see a full picture of what's going on without CLI queries
12:57 boris-42 yfried: hm
12:57 yfried boris-42: shouldn't be too hard to generate a web version of it using the existing "info" queries
12:58 boris-42 yfried: ah so having "Plugins Reference"
12:58 yfried boris-42: yes
12:58 boris-42 yfried: so I am not big expert in sphinx
12:58 boris-42 yfried: but there is the way to add custom python code that generates doc pages
12:58 yfried boris-42: also, I, personally, am still unclear about the terminology re plugins/benchmarks/scenarios/tasks
12:59 boris-42 plugins - everything
12:59 yfried boris-42: I'm no sphinx expert either. I am just thinking out loud
12:59 yfried boris-42: define everything
12:59 boris-42 yfried: so if you make a patch that adds plugins to docs
12:59 boris-42 yfried: that will be nice
13:00 boris-42 but somebody should find time how to do it properly with sphinx
13:00 boris-42 i don't have at the moment enough time
13:00 yfried boris-42: as I said - not a real sphinx guy. who's the rally guy that does that? I'll ask him
13:00 boris-42 yfried: msdubov, but he will be quite busy
13:00 yfried boris-42: let's add that to road map at least
13:00 boris-42 yfried: by other stuff
13:01 boris-42 yfried: it will be low priority for now
13:01 yfried boris-42: sure, but at least it will have some priority
13:02 jjmb joined #openstack-rally
13:03 boris-42 yfried: added
13:03 boris-42 yfried: so btw could you make that small task /
13:03 boris-42 ?
13:03 boris-42 yfried: "--os-endpoint"
13:04 boris-42 yfried: to be recognized by "rally deployment create --fromenv" ?
13:05 yfried boris-42: could try to explain again? I'm not sure I understood what it is you are looking for
13:05 boris-42 yfried: few days ago I asked you can you help me with improving "ExistingEngine" plugin
13:05 yfried boris-42: to search for the os-endpoint env-var and add it to the db upon deployment creation?
13:05 boris-42 yfried: some kind of that
13:06 openstackgerrit Merged stackforge/rally: Added 'alias' decorator  https://review.openstack.org/163861
13:06 boris-42 yfried: like here https://github.com/stackforge/rally/blob/master/rally/cmd/commands/deployment.py#L62-L67
13:06 yfried boris-42: is it just this single var, or are you looking for something more general?
13:06 boris-42 yfried: at least that single var
13:06 boris-42 yfried: because it's crucial
13:07 boris-42 yfried: in ideal it will be nice to accept most of args from env variables like keystone
13:07 boris-42 but other arguments can be done later
13:13 nmagnezi joined #openstack-rally
13:13 yfried boris-42: IIUC os-endpoint is the service-specific endpoint to auth against
13:14 yfried boris-42: and it's different for each service
13:14 boris-42 yfried: for now we are supporing only specific for keystone
13:15 boris-42 yfried: it covers most cases for now
13:15 yfried boris-42: I'm sorry. I don't understand
13:15 yfried boris-42: AFAIK you auth using the 4-tuple (name,pass,tenant, url)
13:16 yfried boris-42: and you get the service cataloge from keystone and a token
13:16 yfried boris-42: the service cataloge holds ALL endpoints for all services
13:17 yfried boris-42: Unless I'm mistaken, there's no use of having only keystone-endpoint without having a way to use tokens as well
13:17 yfried boris-42: or any endpoint
13:18 boris-42 yfried: there is
13:18 boris-42 yfried: without specification of keystone os-endpoint you can run rally against Mirantis OpenStack
13:19 boris-42 yfried: because you don't have access to the managemnt network
13:19 yfried boris-42: "can't"?
13:19 boris-42 yfried: setting os-endpoint allows you to override value from keystone catalog
13:19 boris-42 yfried: yep you can't
13:20 yfried boris-42: but is will set the kestone endpoint for all other clients in deployment and they will not work, because they have different endpoint
13:20 yfried s
13:21 boris-42 yfried: believe me
13:21 boris-42 yfried:  it works well
13:21 boris-42 yfried: https://github.com/stackforge/rally/blob/master/samples/deployments/existing-with-given-endpoint.json#L6
13:21 boris-42 yfried: ^ it is already supported
13:22 boris-42 yfried: we need just to improve --fromenv case
13:22 boris-42 yfried: to accept it that is all
13:22 yfried boris-42: sorry for arguing. I'm still trying to understand the goal here
13:22 yfried boris-42: so it works with file
13:22 boris-42 yfried: ?
13:22 boris-42 yfried: if you try to access MOS not from controller
13:23 yfried boris-42: ^ and you can't do that with with file?
13:23 boris-42 you don't have access to managemnt netowrk (and managment url) => so you are not able to create any users/tenants/ => Rally doesn't work
13:23 boris-42 yfried: we CAN do
13:23 boris-42 yfried: but we can't do that from env
13:23 boris-42 yfried: what I am asking to add abbility to do it from env
13:23 yfried boris-42: I'm just concerned that pulling endpoint from env isn't regular usage of openstack
13:24 yfried boris-42: ok I'll add it
13:24 boris-42 yfried: it is
13:24 yfried boris-42: but no as required. just optional
13:24 yfried boris-42: is there a bp/bug in launchpad I should reference?
13:24 boris-42 yfried: https://github.com/openstack/python-keystoneclient/blob/master/keystoneclient/shell.py#L168-L174
13:25 boris-42 yfried: it is regular usage of keystoneclient https://github.com/openstack/python-keystoneclient/blob/master/keystoneclient/shell.py#L168-L174
13:25 boris-42 yfried: rally doesn't support it and it produce issues
13:25 yfried boris-42: ^ this is when using keystone client specifically.
13:25 boris-42 yfried: rally as well uses keystoneclient
13:26 boris-42 yfried: regular kesytoneclient
13:27 boris-42 yfried: and we should be able to pass this argument when we are creating deployment --fromenv
13:27 yfried boris-42: but we'll have to add it to DB as keystone-endpoint
13:27 boris-42 yfried: it is already added
13:27 boris-42 yfried: everything is already DONE
13:27 boris-42 yfried: you should just pass it from env -> to plugin
13:27 yfried boris-42: ok
13:27 boris-42 few lines of code
13:28 yfried boris-42: asking again about lauchpad
13:29 boris-42 yfried: it's tooooooo small change for bp
13:29 boris-42 yfried: but if you wish you can add bug for it
13:29 boris-42 yfried: https://github.com/stackforge/rally/blob/master/rally/cmd/commands/deployment.py#L96 just one line here
13:30 boris-42 endpoint:  os.environ.get("OS_ENDPOINT")
13:30 boris-42 and probably changing the documentation
13:33 jjmb1 joined #openstack-rally
13:33 boris-42 yfried: what I think is that maybe we should pass more arguments of keystone clinet
13:33 boris-42 at least all that we can accept by current existing engine plugin
13:38 openstackgerrit Yair Fried proposed stackforge/rally: [WIP] Adds endpoint to --fromenv deployment creation  https://review.openstack.org/166607
13:38 yfried boris-42: ^
13:39 yfried boris-42: btw, why am I seeing all sorts of tags on commit messages?
14:05 boris-42 yfried: because guys thinks that it simplifies life
14:05 boris-42 think*
14:26 yfried boris-42: didn't we agree to spec this?
14:28 yfried boris-42: one more q about rally server
14:28 yfried boris-42: I know it can compare 2 tempest runs
14:28 yfried boris-42: but can it compare 2 (or more) rally tasks?
15:08 exploreshaifali joined #openstack-rally
15:41 baker joined #openstack-rally
15:50 jjmb joined #openstack-rally
16:12 tehbrut joined #openstack-rally
16:37 baker joined #openstack-rally
16:54 exploreshaifali joined #openstack-rally
17:33 yfried joined #openstack-rally
17:44 jjmb1 joined #openstack-rally
17:56 boris-42 yfried: it's not done yet
17:57 boris-42 yfried: it's in road map
18:03 PerfBeing joined #openstack-rally
18:51 PerfBeing joined #openstack-rally
18:58 PerfBeing joined #openstack-rally
19:21 PerfBeing joined #openstack-rally
20:22 jjmb joined #openstack-rally
21:20 andreykurilin_ joined #openstack-rally
21:23 psd joined #openstack-rally
21:27 e0ne joined #openstack-rally
21:42 yfried|afk joined #openstack-rally
21:44 baker joined #openstack-rally
22:04 baker joined #openstack-rally
22:11 mwagner_lap joined #openstack-rally
22:46 rook_ joined #openstack-rally
22:47 psd joined #openstack-rally
23:18 baker joined #openstack-rally

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