| *** dasanind_ has joined #openstack-valence | 08:54 | |
| *** lin_yang has quit IRC | 09:01 | |
| *** dasanind has quit IRC | 09:01 | |
| *** dasanind_ is now known as dasanind | 09:01 | |
| *** lin_yang has joined #openstack-valence | 10:51 | |
| *** openstackgerrit has joined #openstack-valence | 12:15 | |
| openstackgerrit | Akhil jain proposed openstack/valence master: Fixing installation for Ubuntu 16.04 https://review.openstack.org/461369 | 12:15 |
|---|---|---|
| *** hubian has joined #openstack-valence | 14:13 | |
| *** ntpttr_laptop has joined #openstack-valence | 14:56 | |
| *** ananth_n has joined #openstack-valence | 14:56 | |
| ananth_n | #topic valence | 14:57 |
| *** shuquan has joined #openstack-valence | 14:57 | |
| ntpttr_laptop | ananth_n: I don't think we have the meeting bot | 14:58 |
| ananth_n | yes :) i didn't mean to hit 'enter' but it happened | 14:58 |
| shuquan | hi guys | 14:58 |
| hubian | hello all | 14:59 |
| ntpttr_laptop | hi everyone | 14:59 |
| ananth_n | hi shuquan, hubian | 14:59 |
| mkrai | Hi | 14:59 |
| shuquan | long time no see, ananth :) | 14:59 |
| lin_yang | Hello everyone | 14:59 |
| ananth_n | there's still a couple of minutes, so let's see if anyone else joins | 14:59 |
| ananth_n | shuquan yes :) been busy with a few things. | 15:00 |
| ananth_n | ok - i guess we can start | 15:01 |
| ananth_n | madhuri & i had one agenda item. are there others? | 15:01 |
| *** ramineni_ has joined #openstack-valence | 15:02 | |
| ntpttr_laptop | I think some of the things that anusha was emailing me about would be good to talk about | 15:02 |
| ntpttr_laptop | I CCed a few more folks in that but not everyone here | 15:02 |
| ntpttr_laptop | wrt storage APIs | 15:02 |
| *** jfding has joined #openstack-valence | 15:02 | |
| hubian | seems the mutli-podm talking could be a topic | 15:02 |
| ntpttr_laptop | also, just to let folks know, I've gotten the okay to spend some more of my time working on this project again, there was a while that I was all focused on another but now I'll be spending time on Valence again | 15:03 |
| hubian | Hello jianfeng @jfding | 15:03 |
| jfding | hi~ | 15:03 |
| ntpttr_laptop | so you'll see me around more and I'll spend more time w/ reviews and things | 15:03 |
| shuquan | another topic could be the 2.1 api support | 15:03 |
| ananth_n | ntpttr_laptop that is great | 15:03 |
| lin_yang | sound great ntpttr | 15:03 |
| ananth_n | ok - shall I begin with the topic that Madhuri & I had | 15:03 |
| ntpttr_laptop | ananth_n: we can start with your and madhuri's item | 15:03 |
| ananth_n | thanks nate | 15:04 |
| jfding | @ntpttr so great | 15:04 |
| ananth_n | we've been exploring some scalability related items, and it also relates to some extent to multi-podm support | 15:04 |
| ananth_n | because we need to add a 'scheduler' | 15:05 |
| *** Mrittika has joined #openstack-valence | 15:05 | |
| hubian | yeah, the scheduler sould be "must to have" | 15:05 |
| Mrittika | Hi. | 15:05 |
| ananth_n | i've submitted blueprints for scheduler and Mrittika had submitted one on task manager | 15:05 |
| ntpttr_laptop | I think it can definitely be something simple to start | 15:05 |
| ananth_n | that's the ones I'm refering to in this context | 15:05 |
| ntpttr_laptop | I think in most cases people will likely rather select pods themselves | 15:06 |
| ntpttr_laptop | the scheduler being there for backup | 15:06 |
| Mrittika | Did folks have a chance to seethe new blueprints | 15:06 |
| ntpttr_laptop | I'll take a look today, do you have a link? | 15:06 |
| ananth_n | ntpttr_laptop that's how we have it today. but we'd want to move that logic into the scheduler | 15:06 |
| lin_yang | make sense to add scheduler module, will take a look the bp today | 15:07 |
| ntpttr_laptop | ananth_n: move which logic? | 15:07 |
| ntpttr_laptop | selecting pods? I think we should keep the ability to choose a pod when composing | 15:07 |
| lin_yang | ananth: the same question with ntpttr | 15:07 |
| ananth_n | "people will select pods themselves" to "valence takes that decision" | 15:07 |
| Mrittika | Yup | 15:08 |
| ntpttr_laptop | Well Valence might do it as a backup, but I do think we shouldn't take away the ability to make the decision yourself | 15:08 |
| hubian | We could define a file to store all kinds of "scheduler functions for pod manager" , also we need both support munally select by user and auto select by scheduler | 15:08 |
| ntpttr_laptop | I feel like control over which pod you're composing in is important | 15:08 |
| ananth_n | i'll let Mrittika chip in, she also had some points on this | 15:08 |
| ramineni_ | ananth_n: i think both the options are to be provided | 15:08 |
| Mrittika | Good point @ntppttr | 15:08 |
| mkrai | Or we can have both, but the default one would be scheduler making the decision | 15:08 |
| shuquan | agree, there should be multiple scheduler can be configured | 15:09 |
| hubian | yeah ~ | 15:09 |
| ananth_n | we could assume that the scheduler mechanism is the primary, while "user driven" is the exception | 15:09 |
| ntpttr_laptop | sure, just an option in the compose API for selecting a pod, and if it's left blank the scheduler makes the decision | 15:09 |
| ntpttr_laptop | makes sense | 15:09 |
| ananth_n | to second what mkrai says | 15:09 |
| shuquan | and use some scheduler hit to make the decision by the user | 15:09 |
| lin_yang | if user don't specify any pod, scheduler will decide one | 15:09 |
| Mrittika | Option can be provided for both. Add that to the blueprint | 15:09 |
| ramineni_ | ntpttr_laptop: +1 | 15:09 |
| mkrai | ntpttr_laptop: +1 | 15:10 |
| ananth_n | ok - so i guess there is agreement there. that was fast! | 15:10 |
| hubian | ntpttr_laptop +1 | 15:10 |
| ananth_n | :) | 15:10 |
| ntpttr_laptop | link to BP? | 15:10 |
| ananth_n | one sec | 15:10 |
| Mrittika | So here is the use case green field and brown field | 15:10 |
| ntpttr_laptop | I want to open it up in a tab so I get to it today | 15:10 |
| hubian | that's actually the original design | 15:10 |
| ananth_n | https://blueprints.launchpad.net/openstack-valence/+spec/valence-multipodm-scheduler | 15:10 |
| Mrittika | Green field use case valence installs and pods register with it | 15:10 |
| ntpttr_laptop | sweet thanks | 15:11 |
| Mrittika | Brown field you roll in a rack or new servers in an existing setup | 15:11 |
| ramineni_ | i guess hubian already have one implementation in his mult-podm patch? | 15:12 |
| Mrittika | So discovered pods could be listed on the ui | 15:12 |
| ntpttr_laptop | yeah I'm guessing if Valence is already up and running you'll need to manually add the new pod - and when it's added valence can do a sweep to see what resources it has and add them to the DB? | 15:12 |
| ntpttr_laptop | ramineni_ and I had a bit of email correspondence on this | 15:12 |
| Mrittika | No it registers itself | 15:12 |
| ntpttr_laptop | it's sort of a new topic though | 15:12 |
| ntpttr_laptop | Mrittika: how will a pod know how to register itself with valence? | 15:12 |
| lin_yang | Mrittika: you mean register it into podm or valence? | 15:13 |
| Mrittika | The scheduler blueprint is only scheduler | 15:13 |
| Mrittika | @lin_yang valence | 15:13 |
| ntpttr_laptop | I'm not sure what you're saying Mrittika; the user will need to give the new endpoint and credentials of the new pod | 15:14 |
| hubian | yeah , i have the patch for mutli-podm : https://review.openstack.org/#/c/445360/ | 15:14 |
| ntpttr_laptop | how will Valence auto-add new pods that happen to come up | 15:14 |
| mkrai | I think admin has to do that part | 15:14 |
| mkrai | Add new podmanager to valence | 15:14 |
| ntpttr_laptop | me too mkrai | 15:14 |
| Mrittika | Will be adding a blueprint | 15:14 |
| ntpttr_laptop | and then when the new podm is added, valence will automatically discover all the resources it controls | 15:14 |
| hubian | +1 mkrai | 15:15 |
| Mrittika | Can be manual at first | 15:15 |
| ananth_n | we're considering the option of auto discovery of podms. that part requires PODM to be modified, and is a future path | 15:15 |
| lin_yang | +1 ntpttr valence will discover all resources after podm added | 15:15 |
| ananth_n | currently we have a manual method | 15:15 |
| ntpttr_laptop | hmm I'm not sure that I agree that the podm software will want to have logic for telling valence about itself | 15:15 |
| ntpttr_laptop | I think it does make more sense for valence to be doing that work, will the folks writing the pod manager even want that stuff to be in the podm? | 15:16 |
| Mrittika | do add a blueprint on that | 15:16 |
| ananth_n | ntpttr_laptop, that we have to influence them to do :) | 15:16 |
| mkrai | Yes we might in future have some polling service in valence to auto discover new pods | 15:17 |
| lin_yang | and we might need a periodic task to refresh the resource of every podm, so scheduler can depend on that to make decision | 15:17 |
| ntpttr_laptop | ananth_n: but I don't think we should tbh | 15:17 |
| ntpttr_laptop | I don't think that really belongs on the podm side | 15:17 |
| Mrittika | Going back to the scheduler it should be a simple task scheduler with queues | 15:18 |
| Mrittika | Hi Ian's multipodm can use this scheduler | 15:18 |
| ntpttr_laptop | like in OpenStack in general, say the cinder service, storage services aren't auto-registering with it. The open source solution is provided with info on how to connect to the hardware | 15:18 |
| ntpttr_laptop | Mrittika: that sounds good | 15:18 |
| Mrittika | Sorry hubian | 15:18 |
| ananth_n | folks, just a quick time check, there's one more topic I need to discuss, and 2 other topics, maybe we can discuss this in an email chain. maybe schedule further voice/IRC meetings? | 15:19 |
| hubian | no need sorry @ mrittika | 15:19 |
| lin_yang | ananth: sounds good to me | 15:19 |
| ntpttr_laptop | sure we can move to the next topic | 15:19 |
| ananth_n | Mrittika : you wanted to discuss networking | 15:19 |
| shuquan | move forward | 15:20 |
| ntpttr_laptop | I'd like to talk more about this podm auto-register sometime though | 15:20 |
| Mrittika | Yup | 15:20 |
| Mrittika | And storage | 15:20 |
| hubian | I thouth i put a such big change in only one patch , that make these thing more stuff | 15:20 |
| Mrittika | Which one do we need to implement first | 15:20 |
| shuquan | storage +1 | 15:20 |
| Mrittika | Ok hubian | 15:20 |
| ananth_n | @ramineni_ i'd sent you a note on networking | 15:20 |
| ntpttr_laptop | Mrittika: Can we do both in parallel? I'm happy to do storage | 15:20 |
| ntpttr_laptop | I have the existing patch that I can update | 15:20 |
| Mrittika | Ok ntpttr | 15:21 |
| ramineni_ | ananth_n: yes, i dont have particular usecase for networking , but im interested in managing all devices in general though | 15:21 |
| Mrittika | We want a network service and a storage service | 15:21 |
| hubian | I can continue to make up the codes to support mutli-podm in my patch | 15:21 |
| Mrittika | So each service can then do its own thing | 15:21 |
| ramineni_ | Mrittika: can you provide some context, which exact feature you are talking about? | 15:22 |
| Mrittika | The service is used for configuring, get, set compose storage or networki etc | 15:22 |
| hubian | About the redfish and podmanager , seems we do have serval choice @ramineni @ntpttr_laptop | 15:23 |
| hubian | define the compose function in redfish.py or in the podmanager.py ？ | 15:23 |
| lin_yang | Mrittika: besides the valence, new service for network and storage? or new module? | 15:23 |
| ntpttr_laptop | hubian: one sec I don't think we're on that topic yet | 15:23 |
| hubian | oo , it is ok ~ | 15:23 |
| ntpttr_laptop | Mrittika: do you have any other opens on storage/networking? | 15:23 |
| hubian | sorry | 15:24 |
| ntpttr_laptop | hubian: no problem :) | 15:24 |
| Mrittika | Hubian compose can come from podm now but later redfish will have compose | 15:24 |
| ntpttr_laptop | do we have someone willing to work on networking and adding that module? | 15:24 |
| ananth_n | @lin_yang our current codebase uses blocking calls. the suggestion is to break this into services, for scalability. | 15:24 |
| mkrai | Mrittika is suggesting to redesign valence and introduce new RPC servicces to manage compute, storage, networking etc | 15:24 |
| Mrittika | Yes mkrai | 15:24 |
| ananth_n | so we redesign | 15:24 |
| ananth_n | mkrai +1 | 15:24 |
| ntpttr_laptop | mkrai: whoa whoa whoa why would we touch rpc | 15:24 |
| ntpttr_laptop | what's the rationale there, that's a lot of added complexity that we agreed to avoid from the beginning | 15:25 |
| mkrai | for scalabily | 15:25 |
| Mrittika | To make it scalable | 15:25 |
| mkrai | *scalability | 15:25 |
| ntpttr_laptop | at the end of the day all these services are doing is making API calls and DB calls | 15:25 |
| Mrittika | To support multipods | 15:25 |
| ntpttr_laptop | I don't think RPC is going to make it more scalable | 15:26 |
| ntpttr_laptop | RPC is used in openstack for talking between many different major services | 15:26 |
| Mrittika | Change API calls to service based | 15:26 |
| ntpttr_laptop | this would be like cinder adding it's own mini RPC for different storage drivers | 15:26 |
| mkrai | Having just an API service makes valence non scalable | 15:26 |
| Mrittika | Valence needs to be a controller and not driver | 15:26 |
| mkrai | ntpttr_laptop: We could run multiple instances of scheduler, compute, network etc service to scale valence | 15:27 |
| Mrittika | I don't think we need to conclude today. Think about it | 15:27 |
| ntpttr_laptop | okay, I'm open to looking into it | 15:27 |
| ntpttr_laptop | I just remember having this conversation a few months ago w/ Malini and all and deciding that it was overkill | 15:27 |
| ntpttr_laptop | but that was before multi-podm maybe | 15:27 |
| Mrittika | Yes | 15:28 |
| Mrittika | And Malini is not there😀 | 15:28 |
| ananth_n | ntpttr_laptop i agree, we had decided to keep away from few things. but as you mention, multi podm is a new dimension | 15:28 |
| mkrai | Yes we will share an architecture design to all to have a look and provide their input | 15:28 |
| ntpttr_laptop | that doesn't make what she said invalid though ;) | 15:28 |
| mkrai | via email | 15:28 |
| lin_yang | yep, to make sure everyone fully understand the context at first, please drop us a mail or bp | 15:28 |
| ntpttr_laptop | sounds good mkrai | 15:28 |
| Mrittika | Just joking | 15:28 |
| ramineni_ | lin_yang: +1 | 15:28 |
| ananth_n | mkrai and I had a discussion | 15:28 |
| ananth_n | mkrai +1 | 15:29 |
| ananth_n | lin_yang sure | 15:29 |
| hubian | wait for the email :) | 15:29 |
| mkrai | Sure we will do it soon | 15:29 |
| ananth_n | we're almost out of time. | 15:29 |
| hubian | thanks · | 15:29 |
| ananth_n | ntpttr_laptop wanted to discuss the mail thread which anusha had initiated | 15:29 |
| ntpttr_laptop | maybe we need to reserve more time for this meeting in the future? | 15:29 |
| mkrai | yes | 15:30 |
| Mrittika | Yes | 15:30 |
| mkrai | and also have meeting in proper format | 15:30 |
| ntpttr_laptop | we can talk about that over email | 15:30 |
| Mrittika | How about 1hr | 15:30 |
| ramineni_ | ananth_n: ntpttr_laptop: yes | 15:30 |
| ntpttr_laptop | mkrai: +1 | 15:30 |
| lin_yang | agree one hour make more sense | 15:30 |
| mkrai | Chair, meetbot | 15:30 |
| mkrai | etc etc | 15:30 |
| ntpttr_laptop | mkrai: we need to reserve a slot in a meeting channel | 15:30 |
| mkrai | ntpttr_laptop: Yes | 15:30 |
| ntpttr_laptop | I think that would be really helpful | 15:30 |
| shuquan | how about one hour earlier? :) | 15:30 |
| mkrai | let's have an email for that also | 15:30 |
| lin_yang | mkrai: I remember you have enabled the meetbot before, right? | 15:31 |
| Mrittika | Ok bye then | 15:31 |
| ntpttr_laptop | shuquan: half hour earlier? | 15:31 |
| ntpttr_laptop | we can talk about this on email, I have another meeting :) | 15:31 |
| shuquan | it will be great | 15:31 |
| ntpttr_laptop | thanks everyone | 15:31 |
| shuquan | thanks | 15:31 |
| lin_yang | thanks everyone | 15:31 |
| hubian | half hour earlier +1 | 15:31 |
| mkrai | lin_yang:I think it was there in our infra patch | 15:31 |
| ananth_n | bye folks. another meeting to jump | 15:31 |
| mkrai | But not sure whether we have one or not. | 15:31 |
| ramineni_ | bye all | 15:31 |
| shuquan | bye | 15:31 |
| mkrai | Bye. Thank you all | 15:31 |
| *** ramineni_ has left #openstack-valence | 15:31 | |
| ntpttr_laptop | bye | 15:31 |
| *** shuquan has quit IRC | 15:32 | |
| *** ananth_n has quit IRC | 15:32 | |
| hubian | @ntpttr @ramineni do you have time to have a talk for the emails on redfish and podm ? | 15:33 |
| hubian | ok , let's continuing this talk by emails | 15:34 |
| ntpttr_laptop | hubian: sorry I'm in another meeting now | 15:34 |
| *** hubian has quit IRC | 15:35 | |
| *** Mrittika has quit IRC | 15:35 | |
| *** ntpttr_laptop has quit IRC | 16:56 | |
| *** ntpttr_laptop has joined #openstack-valence | 17:34 | |
| *** ntpttr_laptop has quit IRC | 17:35 | |
| *** jfding has quit IRC | 19:51 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!