This repository has been archived by the owner on Jul 16, 2020. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 51
Weekly Meeting 2016 06 23
Tim Pepper edited this page Jul 7, 2016
·
3 revisions
- opens (5 minutes)
- bugs (10 minutes)
- Testing strategy
meeting bot had an issue...here's a copy/paste of a local irc client's logs:
ciaomtgbot: Meeting started Thu Jun 23 15:02:35 2016 UTC. The chair is AmyLee7. Information about MeetBot at http://wiki.debian.org/MeetBot.
ciaomtgbot: Useful Commands: #action #agreed #help #info #idea #link #topic.
AmyLee7: o/
tcpepper: o/
sameo: o/
chamings: o~
AmyLee7: #topic opens
leoswaldo: o/
tcpepper: semi-open: imho we should plan in the coming weeks to decide our path wrt OpenStack.
albertom: o/
leoswaldo: I have a couple
tcpepper: earlier June happens tend to preclude us from becoming an official project
sameo: tcpepper: Should we start discussing that today ?
tcpepper: we could or give it a week to kick around in people's minds and form an opinion to then share next week?
tcpepper: it's a weighty topic
markusry: sameo: No we don't. https://github.com/01org/ciao/wiki/Gatekeeper-schedule
sameo: tcpepper: Ok. Could you give us the deadline ?
sameo: markusry: Fair enough...
tcpepper: sameo: I'll put it on next week's agenda if folks agree at least this week that we should drive to a decision
sameo: markusry: Now I understand why you were expecting me to merge 305 :)
sameo: tcpepper: Sounds good.
markusry: sameo: RIght.
tcpepper: I wasn't sure where folks were on the topic...we could just leave it on the table for a while yet
AmyLee7: tcpepper: agree and we should also make sure Arjan is in attendance as I know he has been discussing with management here too
AmyLee7: maybe we bring it up in our discussion Monday?
sameo: tcpepper: I think it's time for us to seriously think about it.
sameo: mcastelino: http://lwn.net/Articles/691887/
tcpepper: also the nova mid-cycle is coming to Oregon in a couple weeks. so a decision by then would drive whether we participate in that
tcpepper: #action tcpepper add OpenStack big tent discussion to next week's agenda and bring in Arjan
AmyLee7: any other opens?
leoswaldo: I have two
AmyLee7: ok go ahead leoswaldo
leoswaldo: as you might know I've been working on rally-ciao
leoswaldo: rally now can boot/delete machines,
sameo: leoswaldo: On top of ciao...
mcastelino: leoswaldo: VMs and containers?
leoswaldo: there are still some other call I need to create but they are informational, so if we test a boot it should work, it will through 4 exceptions
leoswaldo: for which I opend issues #299 #300 #301 #302
markusry: Does rally have fixed set of tests that it runs through?
leoswaldo: mcastelino: only tested with vms
sameo: #action tcpepper to add openstack path discussion to next week agenda.
leoswaldo: I would need AmyLee7, sameo to set the priority for this issues, I consder they are low like P2/P3 since they are only informational
leoswaldo: markusry: yes, rally calls them scenarios
sameo: leoswaldo: Those missing endpoints are P2 imho.
markusry: how many can it run now on top of ciao?
leoswaldo: for boot/delete and listing
AmyLee7: #action Amy assign priority (P2) to issues #299 #300 #301 #302
leoswaldo: the top pririty was boot/delete so I would need to talke to same to see what are going to be our next priority
sameo: #info rally on top of ciao can boot, delete and list instances
leoswaldo: I have only tested on a ciao vm cluster, so the numbers I've got are not nice, because of my hardware
markusry: Great. Any idea what percentage of rally's functionality this is?
leoswaldo: but I'm planning to make a run on bare metal the begginingnext week
sameo: #info leoswaldo to run rally on top of ciao on bare metal cluster next week.
leoswaldo: no to be sincere. i can make a research to see our coverall, and bring a status next week
obedmr: markusry: we have the most important one, here's the full list https://github.com/openstack/rally/blob/master/rally-jobs/nova.yaml
obedmr: _ones
sameo: obedmr: But which ones of those do we support ?
markusry: I see. Thanks
AmyLee7: #action leoswaldo will discuss status of rally on ciao bare metal next week
obedmr: sameo: boot, destroy, list
leoswaldo: vms*
tcpepper: we wont support a lot of that list yet as it includes things like pause/unpause, attach storage, &tc.
AmyLee7: #info full list of rally functionality is here- https://github.com/openstack/rally/blob/master/rally-jobs/nova.yaml
tcpepper: that list though does amount to a nice gap list
markusry: It looks like there are about 70 test cases there
sameo: leoswaldo: Maybe you could propose which next scenarios you think we should support next ?
mcastelino: tcpepper: would we need to pass all of them to get into openstack by next OSS?
tcpepper: mcastelino: I don't think so. "get into" is more about following process. if we passed all of those by October that would be amazing :)
leoswaldo: sure, I'll take a look and make a priority list on importance
tcpepper: (amazingly awesome ;)
AmyLee7: #action leoswaldo will go through the functionality and create a prioritized list
leoswaldo: I think that's it from my opens
AmyLee7: ok thank you leoswaldo!
markusry: I'd just like to raise the workloads issue
markusry: i.e., that we should spend some time looking at ciao from the user's perspective
markusry: understand how to create cloud apps on top of it, etc.
markusry: Make sure it's easy
AmyLee7: markusry is this the feature you added to sameo feature deck?
tcpepper: we need a plan for tooling around workload generation, and somebody to actually make some
markusry: Maybe another agenda item for next week.
markusry: Yes
markusry: And to get a good idea of what other orchestrators do an offer here.
obedmr: Maybe Heat orchestration could help with
AmyLee7: yes I think we should discuss and definitely spend time in the next week doing some research
tcpepper: the long goal was definitely to be able to run some existing heat template
markusry: And I think that's the key. We need to spend some time looking at it like we did with storage
markusry: RIght now I have this vague suspicion that something is missing but I'm not quite sure what
sameo: markusry: Looking at kubernetes would be handy as well.
AmyLee7: markusry: do you want to take the lead on putting together a proposal to review?
tcpepper: but we'd also talked about some semi-synthetic workloads like "bitcoin miner" or "wordpress blog". for the later we could even build (if one doesn't exist) a tool to do random greeking-in of pseudocontent.
markusry: Yes.
tcpepper: #link http://www.lipsum.com/
obedmr: I could start doing some research on Heat for CIAO
obedmr: what would be the plan for this orchestration thing, is it planned to have on Sprint 1, 2 or 3?
sameo: #action markusry to build a workload definition/building tooling proposal
markusry: and ultimately some worklods
sameo: obedmr: sprint2...Amy ?
albertom: we really need a way to assign floating ips to vms
markusry: workloads
AmyLee7: yes sprint2
obedmr: ok
sameo: #action obedmr to run some research on Heat for ciao
obedmr: sure
albertom: currently the vms are a bit "useless" since services installed there have to be port forwarded twice
albertom: on cnci and hosts
mcastelino: albertom: tht is not true
mcastelino: they only need to forwarded once on the CNCI
albertom: oh, well anyways, you cannot have two vms serving different webpages on the same port :P
mcastelino: albertom: that is a different problem, it is not about port forwarding twice
albertom: yeah sorry for stating it bad in the first time
mrkz: mcastelino: maybe an example/blog/tutorial on how to archive albertom's user case would be appreciated from the user perspective :)
mcastelino: albertom: and the way you achieve this is typically usinga LBR. Also can you expand the use case offline. I do not understand
albertom: ok will talk about it later, lets continue
tcpepper: bug time?
AmyLee7: yes
AmyLee7: #topic bug scrub
AmyLee7: No P1's :)
tcpepper: AmyLee7: based on what query?
AmyLee7: #link first P2 opened April 8, https://github.com/01org/ciao/issues/16
AmyLee7: based on github
tcpepper: https://github.com/01org/ciao/labels/P1
AmyLee7: those are not bugs
tcpepper: doh indeed
tcpepper: https://github.com/01org/ciao/issues?utf8=%E2%9C%93&q=is%3Aopen%20label%3Abug%20label%3AP1
tcpepper: No results matched your search
markusry: #16 haven't started working on yet. The target is for the end of next week. Not sure I'll get it done. I think I have 9 issues to close for next week
tcpepper: re: issue #16 ...is the payload truly restrictions? or requested resource minimums?
markusry: Although maybe it won't be that hard
tcpepper: markusry: could you ``` paste ``` a sample payload into the issue?
markusry: Basically, there's nothing we can do about disk.
markusry: At least there wasn't last time I looked.
markusry: But currently, you can't restrict a container to using 2 cpus or 100MB of memory for example
AmyLee7: #action markusry will work on issue #16 next week and will ``` paste ``` a sample payload into the issue
markusry: if that container is launched by launcher.
AmyLee7: ok anything else on this bug?
AmyLee7: #link P2 opened April 8 https://github.com/01org/ciao/issues/17
AmyLee7: markusry: this one is yours as well
markusry: Yes. This one is sort of related.
markusry: It's stuff that never got implemented when I did the initial docker integration with launcher
sameo: markusry: I believe you can limit the IO bandwidth, but not the disk size.
AmyLee7: #info issue #16 and 17 are related
markusry: sameo: Yes last time I looked docker used a fixed 20GB disk
markusry: You can change this with an environment variable but this is not obviously container specific
markusry: or obviously not
AmyLee7: #action markusry will work on issues #16 and 17 this week
AmyLee7: #link P2 opened May 5 https://github.com/01org/ciao/issues/98
AmyLee7: this one is kristen's
markusry: Well, I'm not sure. I have 9 issues to fix for next week and those two I was planning to defer
markusry: And concentrate on the others
markusry: I'll try to get them fixed of course.
sameo: markusry: I'd say 16 and 17 can wait a couple more weeks.
sameo: AmyLee7: Maybe leoswaldo can take #98
AmyLee7: should we push them to Sprint 2?
sameo: As he's getting familiar with the ciao-controller compute code.
AmyLee7: on #98 Kristen said we can push to sprint 2
AmyLee7: I am ok with leoswaldo taking this
tcpepper: if desired I could look at 98. I've been in some of that controller code a bit related to building out testutil/
markusry: AmyLee7: Yes I think so.
leoswaldo: sure I can take it
AmyLee7: #action AmyLee7 will push issue #16 and 17 to sprint 2
markusry: AmyLee7: Thanks
AmyLee7: #action leoswaldo and tcpepper will look at issue #98
AmyLee7: #link P2 opened June 9 https://github.com/01org/ciao/issues/234
AmyLee7: leoswaldo: this one is assigned to you
leoswaldo: yes
AmyLee7: any updates?
leoswaldo: sent the pull request earlier today
leoswaldo: #link https://github.com/01org/ciao/pull/310
sameo: AmyLee7: https://github.com/01org/ciao/pull/310
tcpepper: I'll review PR 310 later this morning
AmyLee7: #info PR was sent for #issue234 today
leoswaldo: nice, thanks tcpepper
AmyLee7: #action tcpepper will review PR 310 later this morning
AmyLee7: #link P2 opened June 16 https://github.com/01org/ciao/issues/267
AmyLee7: markusry: this one is yours
markusry: Yes.
markusry: It's a quick fix but not part of Sprint1 so I haven't been working on it
AmyLee7: ok
AmyLee7: #action AmyLee7 update issue #267 to sprint 2
AmyLee7: #topic testing strategy
tcpepper: I'll try to be brief
AmyLee7: I have to head down to the open forum soon...
tcpepper: I've started some docu
AmyLee7: ok thank you tcpepper
tcpepper: #link https://github.com/01org/ciao/blob/master/testutil/README.md
tcpepper: We've got four levels of testability currently
tcpepper: We've got four levels of testability currently
tcpepper:
tcpepper: 1) unit tests: component specific unit test
tcpepper: 2) test util: ssntp cluster test of a component with testutil/
tcpepper: 3) cluster test virtual: single vm tests
tcpepper: 4) cluster test on hardware: BAT
tcpepper: I'd viewed #2 as a gap, as it really kept ciao-scheduler's coverage down, so I spend time on testutil/
tcpepper: manohar's invested in #3
tcpepper: and kristen in #4
tcpepper: I think that's a solid set of options and coverage
tcpepper: what do ya'll think?
mcastelino: tcpepper: agree
sameo: I agree.
mcastelino: tcpepper: A comment. we should able to run BAT on 3. Will try it when I get the chance.
tcpepper: the difference between 1/2 and 3 is how many components they test. 1 and 2 are component focused. 3 is all components
tcpepper: mcastelino: good point
markusry: Agree too.
tcpepper: and right now BAT is only on a small nuc cluster. we'll want to get more into BAT and BAT on more hardware
tcpepper: ok so then we've done ok for the month w/o fuentess :)
sameo: mcastelino: Could we consolidate the simgle VM tests into BAT ?
tcpepper: and that was all from me on test
sameo: And run the BAT tests by default on single VM ?
mcastelino: sameo: we should be able to
mcastelino: pending any bugs
mcastelino: in my setup
sameo: mcastelino: Then 3) is no longer a set of tests, but rather a test medium.
mcastelino: sameo: yes I will remove my own tests and use BAT
mcastelino: sameo: that way single VM will setup the env and BAT will test it
sameo: mcastelino: Or/and extend the BAT ones if they don't cover your current single VM tests.
sameo: mcastelino: Cool.
AmyLee7: #action sameo and mrcastelino close on single VM way to cover BAT
sameo: Does that make sense to you tcpepper ?
AmyLee7: ok I have to head out
tcpepper: sameo: ACK
AmyLee7: #endmeeting
Development
Architecture