Hi,
During the Smaug weekly meeting on 2015-12-22 1400 UTC, I mistyped the
start key, added a space key before the #startmeeting.
Please help me to upload the attached log to the past meetings.
Thanks,
Eran
(16:00:34) gampel: #startmeeting smaug
(16:00:40) zhonghua-lee: hi
(16:00:43) gampel: Hi everyone
(16:00:53) haomaiwang left the room (quit: Remote host closed the connection).
(16:00:56) gampel: Welcome to the first smug IRC meeting
(16:01:04) zengyingzhe_: Hi
(16:01:30) haomaiwang [[email protected]] entered the
room.
(16:01:30) Guest57122 [[email protected]] entered the room.
(16:01:43) gampel: Who is here zengyingzhe_ , saggi ?
(16:01:54) zengyingzhe_: I'm here.
(16:01:58) zhonghua-lee: and me
(16:02:24) thorst_ left the room (quit: Ping timeout: 255 seconds).
(16:02:46) gampel: Ok, We have a lot on the agenda so lets start
(16:02:48) otter768 [[email protected]] entered the room.
(16:02:54) gampel: #topic Smaug Overview
(16:02:55) tshtilma left the room (quit: Ping timeout: 256 seconds).
(16:02:57) zengyingzhe_: OK
(16:02:57) zhonghua-lee: OK
(16:03:04) Guest57122 is now known as chenzeng
(16:03:16) saggi [[email protected]] entered the room.
(16:03:23) saggi: Hey
(16:03:24) trozet [[email protected]] entered the room.
(16:03:39) gampel: Hi saggi this is our first IRC so are there any question
about the projects goals plan mission statement
(16:03:55) markvoelker [[email protected]] entered the room.
(16:04:13) amrith is now known as _amrith_
(16:05:01) gampel: I just would like focus the IRC meetings on operational
info whats the plan who is doing
(16:05:14) zhonghua-lee: what's my suggestion is we should public our project
plan, do you agree with me?
(16:05:28) gampel: yes and I think that deep dive design should be done in
our IRC channel
(16:05:38) saggi: or on gerrit
(16:05:48) elynn left the room ("Be back later ...").
(16:05:57) weshay_xchat left the room (quit: Ping timeout: 260 seconds).
(16:05:59) saggi: similar to how we are now discussing the API
(16:05:59) gampel: #topic "Proposed Smaug API v1.0"
(16:06:20) gampel: Do we have any open question/comments on the API ?
(16:06:47) peristeri [[email protected]] entered the room.
(16:06:54) zhonghua-lee: yes, need we present the 'provider' to end user?
(16:06:58) thorst_ [[email protected]] entered the room.
(16:07:18) weshay_xchat [~whayutin@2602:306:c53c:9330:2ad2:44ff:feb4:ce31]
entered the room.
(16:07:27) otter768 left the room (quit: Ping timeout: 256 seconds).
(16:07:42) gampel: currently we want the user to select a provider from a list
of providers
(16:08:02) smcginnis: o/
(16:08:11) sdake [[email protected]] entered the room.
(16:08:14) zhonghua-lee: when we think about backup, maybe what i care about is
the resources, not the "provider'
(16:08:25) sdake left the room (quit: Client Quit).
(16:08:33) saggi: you care about the provider since you care about where you
back up
(16:08:39) zhonghua-lee: hi smcginnis
(16:08:41) markvoelker left the room (quit: Ping timeout: 272 seconds).
(16:08:57) sdake [~textual@fedora/sdake] entered the room.
(16:09:00) smcginnis: zhonghua-lee: Hi!
(16:09:05) gampel: welcome
(16:09:12) zhonghua-lee: @saggi yes, i think that the target
(16:09:23) zhonghua-lee: that's the target
(16:09:46) gampel: the provider encapsulate a selection of plugins and a bank
(16:09:49) zhonghua-lee: it's difficult to understand what's a 'provider'
(16:10:11) alexpilotti [[email protected]] entered the room.
(16:10:27) saggi: zhonghua-lee: It's a ProtectionProvider. It encompasses all
aspects of the protection. Where to backup and how to back up.
(16:10:28) zhonghua-lee: what's your opinion?
(16:10:50) zengyingzhe_: Full name is ProtectionProvider. I think it's clear.
(16:11:05) saggi: It provides protection services. Though we are open to better
naming conventions.
(16:11:27) saggi: But only in the next couple of days since we really want to
finalize the aPI
(16:11:28) saggi: API
(16:11:45) yanyanhu left the room (quit: Quit: Leaving).
(16:12:07) zhonghua-lee: yes,i know it's simple API
(16:12:08) markvoelker [[email protected]] entered the room.
(16:12:15) gampel: we could use ProtectionProvider , what do you all think
(16:12:38) smcginnis: gampel: Clarity is better IMO. That sounds good.
(16:12:44) zhonghua-lee: maybe ,we should think it over, if it's neccesarry to
present the concepe
(16:13:30) gampel: zhonghua-lee: i am not sure i understand
(16:13:35) saggi: I'm OK with changing plan to protection_plan and provider to
protection_provider. I'm also partial to clarity over brevity.
(16:13:48) zhonghua-lee: maybe, protectionProvider is helpful to understand
(16:13:49) markvoelker left the room (quit: Remote host closed the connection).
(16:14:04) gampel: #action rename provider API to protectionProvider
(16:14:05) markvoelker [[email protected]] entered the room.
(16:14:28) thorst_ left the room (quit: Remote host closed the connection).
(16:14:46) zhonghua-lee: sounds good
(16:14:50) gampel: Do we have any additional comments on the API
(16:15:21) dkranz [[email protected]] entered the room.
(16:15:26) gampel: #topic Open Issues
(16:15:26) chenzeng: no
(16:15:37) zhonghua-lee: there are many comments from yinwei, it seems she is
not here now
(16:15:39) Raildo [[email protected]] entered the
room.
(16:15:43) gampel: Yinwei are you here ? can you please explain this topic
(16:15:52) Raildo is now known as raildo
(16:17:45) gampel: is there any open issues, i guess we will address yinwei
question in next meeting
(16:18:16) zhonghua-lee: +1
(16:18:19) zengyingzhe_: OK
(16:18:24) gampel: #topic Boilerplate un-started generic infrastructure tasks
(16:18:33) gampel: #link https://etherpad.openstack.org/p/smaug-bootsrap
(16:19:29) gampel: chenzeng: can you share your status
(16:20:22) saggi: chenzeng: ?
(16:20:26) chenzeng: yes
(16:21:08) chenzeng: I want to use "apscheduler" to implement the time trigger
(16:21:25) rossella_s left the room (quit: Read error: Connection reset by
peer).
(16:21:48) rossella_s [[email protected]] entered the
room.
(16:22:12) saggi: chenzeng: Are there any other openstack projects that use it?
(16:22:32) chenying_ [[email protected]] entered the room.
(16:22:34) chenzeng: yes,
(16:22:37) yamamoto left the room (quit: Remote host closed the connection).
(16:22:43) chenying_: Hi
(16:23:00) smcginnis: I've used Quartz scheduler quite a bit in other
languages. Googling for quartz for python points to apscheduler.
(16:23:04) smcginnis: That's probably a good sign.
(16:23:15) chenzeng: DPS of huawei use it to implement the trigger.
(16:23:52) gampel: i think it look like a good plan
(16:24:16) zhonghua-lee: @chenzeng DPS is not a OPenStack project now
(16:25:05) gampel: chenying_: whats the status of the Boilerplate tasks
(16:25:42) chenzeng: zhonghua:yes,but we can use the same technology.
(16:25:59) zhonghua-lee: @chenzeng ok
(16:26:06) saggi: chenzeng: Remember to send a UML of the design of the service
you start writing actual code. We need to make sure it all adds up.
(16:27:11) saggi: chenzeng: I would rather use existing openstack dependencies
if those are good enough. Please check if this library is already tracked by
openstack. If not we will have to make sure it's tracked if we decide to use it.
(16:28:18) gampel: saggi: +1 textual format + image will be best for the
design please send it to spec folder under doc
(16:28:27) saggi: chenzeng: Also try and do it in plantuml if possible since
it's a format we can review on Gerrit and push to git.
(16:28:28) chenying_: gampel: Boilerplate tasks? You mean Schedule service
base framework?
(16:28:45) chenzeng: saggi:yes, we should reach an agreement.
(16:28:57) toscalix left the room (quit: Read error: Connection reset by peer).
(16:28:57) gampel: no all the task list we have in the etherpad
(16:28:58) saggi: chenzeng: Boilerplate tasks
https://etherpad.openstack.org/p/smaug-bootsrap
(16:29:29) saggi: The stuff we need to get all the infrastructure working
(16:29:32) chenzeng: saggi:ok
(16:29:53) gampel: if someone wants to take a task from that list please
update the etherpad with the status
(16:30:00) saggi: Everyone please put your name after the task you are taking
so we don't step on each other's toes.
(16:30:19) smcginnis: apscheduler is not in global requirements:
https://github.com/openstack/requirements/blob/master/global-requirements.txt
(16:30:32) smcginnis: And it's an MIT license, not Apache. Not sure if that
will be an issue or not.
(16:31:08) piet [[email protected]] entered the room.
(16:31:19) gampel: do anyone know an alternative library ?
(16:31:51) emsomeoneelse left the room (quit: Quit: Connection closed for
inactivity).
(16:31:56) saggi: chenzeng: Please check if an alternative is in the
global-requirements and if not work on pushing apscheduler.
(16:31:58) chenzeng: maybe we should find another one.
(16:32:09) smcginnis: Oh, I do see other packages marked as MIT in the global
requirements. It's probably OK then
(16:32:25) zhonghua-lee: hi, saggi,maybe, we need more description of each task
in etherpad
(16:32:34) smcginnis: It would be good to see if there is something similar
already there, but if not, probably not a problem getting it added.
(16:32:57) saggi: gampel: Put an action item for it on chenzeng
(16:32:59) chenzeng: saggi:i will.
(16:32:59) gampel: #action chenzeng look for an alternative in the
global-requirements for the triers
(16:33:11) chenying_: I will update the status of the task in the etherpad.
(16:33:32) gampel: can you give us a brief status
(16:33:39) saggi: zhonghua-lee: What is unclear?
(16:34:21) saggi: zhonghua-lee: Those are very basic tasks. Just getting the
services and DB running and all the devstack bits.
(16:34:41) chenying_: DB models : The patchs have been submitted.
(16:34:44) zhonghua-lee: @saggi, it's just my suggestion :)
(16:35:26) pnavarro|lunch left the room (quit: Ping timeout: 245 seconds).
(16:35:30) gampel: zhonghua-lee: no problem just add in the etherpad where you
think that we should add more info
(16:35:32) saggi: chenying_: I saw, I'll get to review those soon. I'm still
finding my way around the Gerrit redesign :) .
(16:35:33) adiantum1 [[email protected]] entered the room.
(16:35:43) saggi: gampel: +1
(16:35:51) zhonghua-lee: @gampel ok, i will
(16:36:02) gampel: #topic Scheduler service design status
(16:36:10) gampel: Saggi can you please share your plans with regards
Scheduler design
(16:36:18) chenying_: Scheduler service: base service class has been done.
smaug-scheduler base service will be done this week.
(16:36:46) ihrachys left the room (quit: Quit: My Mac has gone to sleep. ZZZzzz
).
(16:37:06) gampel: as saggi said before we need to make sure we send the design
spec before the coding
(16:37:21) chenzeng: i agree
(16:37:47) adiantum left the room (quit: Ping timeout: 240 seconds).
(16:37:47) adiantum1 is now known as adiantum
(16:37:51) dane left the room (quit: Ping timeout: 246 seconds).
(16:37:58) saggi: I will try and get something up today. We need to remember
that the scheduler service should be replaceable similar to other schedulers in
openstack.
(16:39:01) saggi: Even by third party. Some enterprises might just want to use
Smaug's low level APIs and do all the scheduling policy and tiering themselves.
(16:41:05) gampel: the main blocks are the Trigger engine , the Operation Exe
and how to handle high availability
(16:41:52) zengyingzhe_: saggi, do you consider the scenario that schedule
service crashed?
(16:42:36) zengyingzhe_: This is about HA, like gampel said.
(16:42:41) saggi: Yes, it's very important to handle that since it handles high
level operations like multiple deletion
(16:42:43) dslev [[email protected]] entered the room.
(16:43:11) zhonghua-lee: is it too early to think about HA?
(16:43:34) saggi: zhonghua-lee: I think we should keep it in mind. We might not
implement it for V1 but we should know that we can.
(16:43:50) gampel: saggi +1
(16:43:58) smcginnis: I can say thinking about it up front will be a lot easier
than trying to fit it in after the fact. ;)
(16:44:10) Qiming left the room.
(16:44:23) gampel: saggi: when do you think you could provide the high level
design block for the Scheduler service
(16:44:35) zhonghua-lee: i think HA is a big solution
(16:45:02) alexpilotti left the room (quit: Remote host closed the connection).
(16:45:17) trozet left the room (quit: Ping timeout: 276 seconds).
(16:45:20) alexpilotti [[email protected]] entered the room.
(16:45:20) gampel: smcginnis: i agree it must be in the design
(16:45:31) smcginnis: zhonghua-lee: Yeah, just keeping it in mind so it's not a
major redesign down the road to add it will help safe some headache.
(16:45:40) smcginnis: s/safe/save/
(16:46:05) zhonghua-lee: smcginnis yeah, i totally agree with you
(16:46:29) saggi: I will have some preliminary UML up tomorrow.
(16:46:36) lixinhui_ left the room (quit: Ping timeout: 246 seconds).
(16:46:51) gampel: #action saggi provide high level design block for the
Scheduler service
(16:47:01) saggi: A base from which we could build on
(16:47:54) gampel: I think yinwei started working on the Protection service
design
(16:48:25) chenying_: yes yinwei is working on Protection service design.
(16:48:28) chenzeng: I have a question, why does the schedule service interact
with DB?
(16:49:03) alexpilotti left the room (quit: Remote host closed the connection).
(16:49:06) saggi: chenzeng: It needs a place to save all the triggers and track
transactions. It might even have it's own DB in the future.
(16:49:09) eharney left the room (quit: Quit: Leaving).
(16:49:13) zhonghua-lee: gampel: yes, but it seems she has a lot of questions
about the relationship between API and Protection service.
(16:49:20) alexpilotti [[email protected]] entered the room.
(16:49:27) spotz_zzz is now known as spotz
(16:49:58) akamyshnikova left the room (quit: Ping timeout: 250 seconds).
(16:50:01) gampel: chenzeng: it is not the same tables as the API/plan and it
could be a different DB as saggi said
(16:50:27) saggi: zhonghua-lee: She can find me on IRC or ask on the API patch.
If there are some stuff that are unclear we would probably want to modify the
documentation so it's clearer anyway.
(16:50:33) chenzeng: ok.
(16:50:43) zengyingzhe_: Is this the decoupling purpose? Use different DB.
(16:51:00) zhonghua-lee: saggi: ok , i will talk to her
(16:51:01) bdrich_ [[email protected]] entered the room.
(16:51:01) chenzeng: it means the schedule service need DB to store its own
information?
(16:51:25) zhurong left the room (quit: Quit: My Mac has gone to sleep. ZZZzzz
).
(16:51:36) saggi: Yes, we would like it not to depend on the API service for
operation.
(16:51:53) gampel: zengyingzhe_: it is completely different data operational
not policy
(16:52:41) zengyingzhe_: I see
(16:52:44) gampel: in the V1 we could use the same DB just different tables all
together
(16:52:52) alkhod_ left the room (quit: Ping timeout: 252 seconds).
(16:52:52) smcginnis: Scheduler is a bit of an overloaded term in this case.
Everywhere else in OS it is as in task placement.
(16:52:54) chenzeng: understand
(16:53:05) smcginnis: In this case it is scheduler as in time scheduler.
(16:53:10) smcginnis: Or perhaps both.
(16:53:18) smcginnis: If I'm making any sense. :)
(16:53:27) pnavarro|lunch [[email protected]] entered the room.
(16:53:33) saggi: smcginnis: It also tracks the tasks and does transactions.
(16:54:18) gampel: smcginnis: i agree we might need to find a better name for
the service but it will do the automatic operation work flow as well
(16:54:23) saggi: smcginnis: For example, if you scheduled to delete all
checkpoints older than 10 days. It will query the protection provider and
issues and track all the deletions.
(16:54:33) pnavarro|lunch is now known as pnavarro
(16:55:12) smcginnis: saggi: True. So just pointing out that for folks coming
from other projects, it may not be clear right away that it is not just
something that takes a request and routes it to where it needs to be executed.
(16:55:20) smcginnis: It is much more in this case.
(16:55:59) smcginnis: So that either just needs to be pointed out or
documented, or it could possibly be called something else and broken out
separately from the "traditional" scheduler concept of other openstack projects.
(16:56:48) trozet [[email protected]] entered the room.
(16:57:01) zengyingzhe_: saggi, you mean schedule service is also a monitor?
(16:57:05) saggi: smcginnis: No one reads the docs :) I will try and think of a
better name. Maybe PlicyEnfocer ?
(16:57:10) gampel: #action saggi, gampel find alternative name for the
scheduler service(ideas are welcome )
(16:57:18) smcginnis: saggi: Very true! ;)
(16:57:40) smcginnis: PolicyEnforcer, TaskManager, ExecutionEngine, ...
(16:57:47) smcginnis: No idea what works best. :-D
(16:57:51) saggi: zengyingzhe_: It will monitor the only the automatic tasks.
Checkpoints that were invoked manually will have to be tracked manually.
(16:58:17) chenzeng: saggi:does the schedule service interact with DB when it
is running?
(16:58:57) saggi: Yes, for example if it failed to run a backup it will have to
make the execution as failed.
(16:59:05) saggi: It will also need to keep state about the transaction
(16:59:14) gampel: we are almost out of time i suggest saggi will have the
design up for review tomorrow and we could continue there
(16:59:23) saggi: gampel: _1
(16:59:24) saggi: +1
(16:59:28) gampel: Please look in the patches Open for review
(16:59:28) chenzeng: +1
(16:59:40) maishsk left the room (quit: Quit: maishsk).
(16:59:45) gampel: thank you every one
(16:59:49) saggi: Just while I still have everyone's attention
(17:00:00) zhonghua-lee: thank you
(17:00:02) saggi: If I missed you comment on the API send it again.
(17:00:05) chenzeng: thanks
(17:00:09) zengyingzhe_: Thanks. Bye.
(17:00:11) saggi: Thanks you everybody
(17:00:20) chenying_: Thanks you all.
(17:00:24) smcginnis: Thanks.
(17:00:32) saggi: and thanks smcginnis for joining. We appreciate you
experience.
(17:00:39) saggi: your
(17:00:43) smcginnis: Glad to!
(17:00:53) haomaiwang left the room (quit: Remote host closed the connection).
(17:00:58) gampel: yes thank you very much please let us know if the time is
good for you
(17:01:04) tshtilma [[email protected]] entered the room.
(17:01:19) smcginnis: Yeah, this works well.
(17:01:24) haomaiwang [[email protected]] entered the
room.
(17:01:41) gampel: bye everyone happy holidays
(17:02:38) gampel: #endmeeting
(17:03:04) hashar [~hashar@mediawiki/hashar] entered the room.
(17:03:06) smcginnis: The bot must be affected by the infra issues too. That's
too bad.
(17:03:26) gampel: :)
(17:03:29) merooney [[email protected]] entered the room.
(17:04:07) _amrith_ is now known as amrith_______________________________________________
OpenStack-Infra mailing list
[email protected]
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-infra
