Perl 6 - the future is here, just unevenly distributed

IRC log for #openstack-rally, 2014-08-17

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

All times shown according to UTC.

Time Nick Message
00:34 shakayumi joined #openstack-rally
01:14 shakayumi joined #openstack-rally
01:34 rook joined #openstack-rally
01:40 shakayumi joined #openstack-rally
01:48 ilbot3 joined #openstack-rally
01:48 Topic for #openstack-rally is now ☁ Rally IRC chat logs http://irclog.perlgeek.de/openstack-rally ☁ Key persons to ask:  boris-42, rediskin, andreykurilin, k4n0  ☁ Open tasks https://trello.com/b/DoD8aeZy/rally ☁ To publish changes to Rally:  https://wiki.openstack.org/wiki/Rally/Develop#How_to_contribute   ☁ Rally weekly updates:  https://wiki.openstack.org/wiki/Rally/Updates
01:53 tzabal left #openstack-rally
02:21 shakayumi joined #openstack-rally
02:31 shakayumi joined #openstack-rally
03:03 shakayumi joined #openstack-rally
04:12 shakayumi joined #openstack-rally
05:07 shakayumi joined #openstack-rally
07:49 AlexF joined #openstack-rally
08:10 asalkeld joined #openstack-rally
08:42 shakayumi joined #openstack-rally
09:23 shakayumi joined #openstack-rally
10:05 shakayumi joined #openstack-rally
11:06 shakayumi joined #openstack-rally
11:50 tzabal joined #openstack-rally
14:46 jaypipes joined #openstack-rally
15:17 shakamunyi joined #openstack-rally
15:21 shakamunyi joined #openstack-rally
15:51 AlexF joined #openstack-rally
17:05 shakamunyi joined #openstack-rally
17:20 rediskin joined #openstack-rally
17:35 AlexF joined #openstack-rally
18:28 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Fix semantic validation of context images  https://review.openstack.org/113904
19:12 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Decrease jobs time in gates  https://review.openstack.org/114839
20:06 rmk How can I get rally to run pdb during a benchmark?
20:06 rmk Seems to ignore it
20:10 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Decrease jobs time in gates  https://review.openstack.org/114839
20:13 boris-42 rmk yep
20:13 boris-42 rmk that's why we have serial runner
20:13 rmk ah
20:13 rmk i hadn't thought of using that, doh
20:13 boris-42 rmk =)
20:14 boris-42 rmk that's why I create osprofiler for openstack=)
20:14 boris-42 rmk http://boris-42.github.io/nova_fun3.html
20:14 boris-42 rmk ^ this is what is happening in nova & glance during nova boot
20:14 boris-42 rmk command from cli
20:14 rmk hah wow
20:15 boris-42 rmk ya quite soon (in Juno)
20:15 boris-42 rmk all python clients will have abbilit
20:15 boris-42 rmk to put in it --profile
20:15 boris-42 rmk and get traces via all services that are invovlide in request
20:15 rmk I must be missing something.  Trying to add a new admin cleanup job and it doesn't seem to be getting run at all.
20:15 rmk I actually don't see any admin cleanup running.
20:15 boris-42 rmk we refactored recently that
20:16 boris-42 rmk now you should specify explicit
20:16 rmk Yeah trying to fix my patch
20:16 boris-42 rmk "clenau" and "admin_cleanup"
20:16 rmk Right
20:16 boris-42 rmk are different context
20:16 rmk I think I see what I missed
20:16 rmk CONFIG_SCHEMA
20:16 rmk I didn't update it
20:17 boris-42 rmk btw are you interested
20:17 boris-42 rmk in work on comparing tasks?
20:18 boris-42 rmk and btw if we make virtual meet up are you going to take a part?
20:18 rmk I'd like to but it might be difficult
20:21 boris-42 meetup or coding?
20:24 rmk meetup
20:24 rmk @base.scenario(context={"cleanup": ["nova"], "admin_cleanup": ["nova"]})
20:25 rmk Can something be part of two cleanups like that?
20:25 boris-42 rmk yep
20:25 rmk All the nova jobs need to be in both
20:25 boris-42 rmk cleanups are different context
20:25 boris-42 rmk so yep
20:25 boris-42 rmk why not?
20:25 rmk It's fine, just wanted to make sure I was using it correctly
20:25 boris-42 rmk we splitter these context as a part of NON admin benchmarking
20:26 boris-42 rmk e.g. when you have 2 LDAP users and it's all=)
20:26 rmk Cool code works now
20:28 openstackgerrit Tzanetos Balitsaris proposed a change to stackforge/rally: Fix semantic validation of context images  https://review.openstack.org/113904
20:28 rmk That was driving me nuts, I couldn't figure out why it wasn't being run at all heh
20:28 boris-42 rmk -(
20:28 boris-42 rmk =)
20:28 boris-42 rmk ya we have a lot of refactroing in rally
20:28 boris-42 rmk we are refactoring since the begging=)
20:29 rmk This is probably the most abstracted code I've seen
20:29 rmk Not necessarily bad, just complex
20:29 rmk Pythonic :)
20:29 boris-42 rmk sorry about that=)
20:29 boris-42 rmk from one side users want flexibility from other we have to keep simplicity
20:29 boris-42 rmk quite hard task=)
20:29 ajayaa joined #openstack-rally
20:30 rmk no doubt
20:30 boris-42 rmk if you know how to make anything there simpler (with keeping functionality) I will be happy+)
20:31 boris-42 rmk hehe soon "rally language" lol
20:38 openstackgerrit Rafi Khardalian proposed a change to stackforge/rally: Disassociate Nova networks during cleanup  https://review.openstack.org/110814
20:39 rmk ^^ Hopefully that's the correct way to do it this time.
20:40 boris-42 tzabal gate-rally-dsvm-rallySUCCESS in 24m 38s
20:40 boris-42 tzabal nice!=)
20:40 tzabal boris-42 yes ^^
20:41 boris-42 tzabal merging=)
20:41 tzabal boris-42 gate-rally-dsvm-neutron-rally took 28m 35s thought
20:41 boris-42 tzabal it depends on nodes as well
20:41 tzabal boris-42 right
20:41 boris-42 tzabal but that one can be as well improved
20:41 boris-42 tzabal in any case=)
20:41 tzabal boris-42 yeap
20:41 boris-42 rmk btw you accidentally remove one benchmark
20:41 boris-42 rmk with your patch
20:42 boris-42 rmk https://review.openstack.org/#/c/113096/5/rally-scenarios/rally.yaml
20:42 boris-42 rmk could you add back (copy paste this one without auto_assing_nics: false )
20:48 rmk oops
20:49 penguinRaider boris-42, in https://review.openstack.org/#/c/113119/11/rally/benchmark/scenarios/requests/http_requests.py if the response is ok or response equal to status code should something be returned?
20:49 boris-42 rmk no worries just 2 benchmarks required now (cause there is option is)
20:49 boris-42 penguinRaider nope why?
20:50 boris-42 penguinRaider you should raise exception in case of failure, and don't raise any exception otherwise
20:50 boris-42 penguinRaider return values are used when there is some statistics for aggregation
20:50 boris-42 penguinRaider but here we don't have (or it's okay, or it is not)
20:53 penguinRaider boris-42, ohh ok thanks
21:04 boris-42 tzabal seems like nova-network is faster then neutron gate!
21:04 boris-42 tzabal seems like one more patch required =)
21:08 boris-42 tzabal so you reduce about 8 minutes ?)
21:08 boris-42 tzabal nice =)
21:08 openstackgerrit A change was merged to stackforge/rally: Decrease jobs time in gates  https://review.openstack.org/114839
21:09 shakayumi joined #openstack-rally
21:14 boris-42 rmk oh that is really bad
21:14 boris-42 rmk https://review.openstack.org/#/c/110814/7/rally/benchmark/scenarios/nova/servers.py
21:15 boris-42 rmk we can't do this ...
21:15 boris-42 rmk I will reply in comment
21:17 rmk boris-42: Not sure how else to deal with that
21:17 boris-42 rmk I am making large comment that should help
21:18 boris-42 writing
21:18 rmk and yeah your logic is better
21:18 boris-42 rmk let's actually discuss here
21:18 boris-42 rmk who is doing associate?
21:19 boris-42 rmk and why it is not cleaned up?
21:19 rmk Nova network associates automatically
21:19 rmk There's a pool of available networks.  If a tenant does not have a network, it will be assigned one when the first instance boots.
21:20 boris-42 rmk but I know that there is perdioc task in nova
21:20 rmk To disassociate networks?
21:20 boris-42 rmk that cleans up these networks
21:20 boris-42 rmk ya
21:20 rmk Are you sure?
21:20 boris-42 rmk lemme find it
21:20 boris-42 rmk otherwise openstack won't work
21:20 boris-42 rmk e.g. users just create and then delete networks => and they are forever associate
21:21 rmk Yep, that's what will happen.
21:21 boris-42 rmk I really don't know why in nova this was done through ass
21:21 boris-42 (I mean periodic task)
21:21 rmk I'm fairly certain there's no periodic job
21:21 boris-42 lemme check twice
21:25 openstackgerrit Rafi Khardalian proposed a change to stackforge/rally: Disassociate Nova networks during cleanup  https://review.openstack.org/110814
21:26 boris-42 rmk oh nova becomes really unreadable
21:26 boris-42 rmk after objects stuff
21:26 rmk yep
21:27 boris-42 2 years before i liked it much more
21:28 rmk Only periodic I see disassociates stale fixed IPs, not entire networks
21:28 rmk boris-42: And yes I agree, I think it's worse now
21:28 openstackgerrit kumar rishabh proposed a change to stackforge/rally: Adding http_requests scenario  https://review.openstack.org/113119
21:29 rmk Lots of code has moved around, very few real problems have been addressed
21:29 boris-42 rmk =(
21:29 boris-42 rmk I have proposal for cinder and others
21:29 boris-42 rmk how to make rolling upgrades
21:29 boris-42 rmk will try to implement in cinder it
21:29 rmk so yeah I don't know how to avoid calling an admin cleanup for every nova scenario
21:29 boris-42 rmk most complicated part are DB migrations
21:30 boris-42 rmk so I know one way
21:30 boris-42 rmk this code should be in another place
21:30 boris-42 rmk with a huge comment: "nova is crap" =)
21:31 rmk lol
21:31 boris-42 rmk https://github.com/stackforge/rally/blob/master/rally/benchmark/context/users.py#L141
21:31 boris-42 rmk ^ during deletion of tenants
21:31 boris-42 rmk this context will be used only if admin is presented
21:32 boris-42 rmk if you already have users you won't care about associated networks
21:32 boris-42 rmk but i won't accept your patch without comment lol =)
21:32 boris-42 rmk I am making collection=)
21:32 rmk haha so literally put it here?
21:33 boris-42 rmk yep
21:33 boris-42 rmk we don't have any other option..
21:33 boris-42 https://github.com/stackforge/rally/blob/master/rally/osclients.py#L47-L48  <- rmk
21:34 rmk i dont mind
21:34 rmk I thought this was ugly too I just didn't see another way to do it via cleanup
21:34 boris-42 rmk if you can run dissociate from users (non admin)
21:34 boris-42 rmk then it can be done in cleanup
21:35 rmk nope, I would have done that if it were possibl
21:35 rmk disassociate is always admin, even if you are the owner
21:35 boris-42 sadness
21:35 boris-42 sumersumertime sadness
21:35 rmk haha
21:36 boris-42 rmk ya benchmarking openstack is hard thing=)
21:36 boris-42 rmk I always laugh when somebody is saying me why not using existing benchmark tools=)
21:44 boris-42 penguinRaider couple of nits more
21:47 penguinRaider boris-42, regarding this https://review.openstack.org/#/c/113119/12/rally/benchmark/scenarios/requests/http_requests.py what if there is ConnectionError during benchmark execution try will handle that ?
21:48 boris-42 penguinRaider nope why?
21:48 boris-42 penguinRaider you'll lose infomration
21:48 boris-42 penguinRaider so if there is some exception => let it go
21:54 openstackgerrit Rafi Khardalian proposed a change to stackforge/rally: Disassociate Nova networks during cleanup  https://review.openstack.org/110814
21:55 rmk boris-42: ^^ is that what you preferred?
21:55 boris-42 http://www.youtube.com/watch?v=m0lHrm70tWw <- this is what I prefer rmk =)
21:55 rmk haha
21:56 boris-42 rmk fu***** green cards=)
21:56 boris-42 rmk 100 gallons / minute
21:56 rmk I was going to say, those things are burning a lot of fuel
21:57 boris-42 rmk I am going to get car someday =)
21:57 boris-42 rmk http://www.youtube.com/watch?v=Drr8R0mtdok
21:57 boris-42 rmk charger 2015 hellcat =)
21:58 boris-42 rmk as well green card=)
21:59 boris-42 rmk 1.5 gallon / minute =)
22:00 boris-42 rmk one issue with your patch..
22:00 boris-42 rmk actually no issue sorry=)
22:01 rmk ah ok cool
22:01 boris-42 rmk but there is tinny hack to make it faster
22:01 boris-42 rmk I 'll left commnet
22:03 boris-42 rmk ok left comment https://review.openstack.org/#/c/110814/9/rally/benchmark/context/users.py
22:04 boris-42 rmk sorry for being so nerd, but this will really make sense if you have 100 tenants
22:08 rmk no big deal, both make sense
22:10 openstackgerrit kumar rishabh proposed a change to stackforge/rally: Adding http_requests scenario  https://review.openstack.org/113119
22:11 openstackgerrit Rafi Khardalian proposed a change to stackforge/rally: Disassociate Nova networks during cleanup  https://review.openstack.org/110814
22:27 AlexF joined #openstack-rally
22:38 asalkeld joined #openstack-rally
23:20 boris-42 rmk some crap in gates
23:20 boris-42 rmk | NovaServers.boot_and_delete_server                            | 0   | max_failure_percent | False   | Maximum failure percent 0% failures, actually 25.0% |
23:20 boris-42 rmk =(
23:20 boris-42 rmk it's not your fault and rally as wel
23:34 rediskin left #openstack-rally

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