Open Stack

Stephen and I from oslo core met the LINE team at Vancouver and discussed
this topic, and in short
we(at least Stephen and I) agreed with the proposal. Let me dump some
topics we discussed there.
# Sorry I thought I posted this somewhere early...

Some concerns were raised about adding consul as a new core component, but
we agreed with using it
as the initial implementation. There were some suggestions to rely on
existing service records stored in DB
but we need some amount of work, to address the following points.
- Not all OpenStack services store service records in DB. (eg ceilometer)
- Even though some services store service records in DB, the available
information there is not enough
 - We have to maintain mapping between host and reachable endpoint
 - Some services spawns worker and need endpoint *per worker*, and we need
endpoint list for all workers

Random port assignment from range is another topic we discussed there, but
we eventually agreed that
this is the required approach, mainly because it's hard to allocate port
numbers for all services running in
a single node(required number of ports can be different based on
services/workers/etc). So we agreed
that the proposed way to dynamically allocate available ports from the
given range is a good solution.

Also, LINE already deployed this feature in their production and it has
been proven to work well in scale.
As we don't see any technical blockers at this moment, we agreed with
moving forward with the proposed
architecture. There can be some improvement like adding the driver
mechanism for cluster backend but,
we would avoid expanding our current scope, until someone is really
interested in working on such topics.

Finally we suggested updating the test plan and adding a devstack(at least
single node) job in the gate
so that we can properly maintain the feature.

Other people might have additional thoughts, but hope the above outcomes
make the current proposal
more clear to all.

Thank you,
Takashi



On Sat, Jun 10, 2023 at 6:14 PM Masahito Muroi <masahito.muroi at linecorp.com>
wrote:

> Hi all,
>
> We have pushed the spec. Please feel free to review it.
> https://review.opendev.org/c/openstack/oslo-specs/+/885809
>
> best regards,
> Masahito
>
> -----Original Message-----
> *From:* "Jay Faulkner"<jay at gr-oss.io>
> *To:* "Julia Kreger"<juliaashleykreger at gmail.com>;
> *Cc:* "Masahito Muroi"<masahito.muroi at linecorp.com>; "Takashi Kajinami"<
> tkajinam at redhat.com>; <openstack-discuss at lists.openstack.org>; "Herve
> Beraud"<hberaud at redhat.com>; "Arnaud Morin"<arnaud.morin at gmail.com>;
> *Sent:* 2023/06/07(水) 07:31 (GMT+09:00)
> *Subject:* Re: [oslo][largescale-sig] HTTP base direct RPC oslo.messaging
> driver contribution
>
> I'm interested in this as well, please add me to the spec if you need
> additional brains :). I'll also be at the summit if you'd like to discuss
> any of it in person.
>
> --
> Jay Faulkner
> Ironic PTL
>
> On Tue, Jun 6, 2023 at 3:14 PM Julia Kreger <juliaashleykreger at gmail.com>
> wrote:
>
> Jumping in because the thread has been rather reminiscent of the json-rpc
> messaging feature ironic carries so our users don't have to run with
> rabbit. I suspect Ironic might be happy to propose it to oslo.messaging if
> this http driver is acceptable.
>
> Please feel free to add me as a reviewer on the spec.
>
> -Julia
>
> On Tue, Jun 6, 2023 at 2:10 PM Masahito Muroi <masahito.muroi at linecorp.com>
> wrote:
>
> Hi,
>
> Thank you everyone for the kindly reply.
>
> I got the PTG situation. Submitting the spec seems to be a nice first
> step.
>
> We don't have public repository of the driver because of internal
> repository structure reason. The repository is really stick to the current
> internal repository structure now. Cleaning up repository would take time
> so that we didn't do the extra tasks.
>
> best regards.
> Masahito
>
> -----Original Message-----
> *From:* "Takashi Kajinami"<tkajinam at redhat.com>
> *To:* "Masahito Muroi"<masahito.muroi at linecorp.com>;
> *Cc:* <openstack-discuss at lists.openstack.org>; "Herve Beraud"<
> hberaud at redhat.com>;
> *Sent:* 2023/06/06(火) 18:50 (GMT+09:00)
> *Subject:* Re: [oslo] HTTP base direct RPC oslo.messaging driver
> contribution
>
> Hello,
>
>
> This is very interesting and I agree having the spec would be the good way
> to move this forward.
>
> We have not requested oslo sessions in the upcoming PTG but Stephen and I
> are attending it so will be
> available for the discussion.
>
> Because some other cores such as Herve won't be there, we'd need to
> continue further discussions after PTG
> in spec review, but if that early in-person discussion sounds helpful for
> you then I'll reserve a table.
>
> Thank you,
> Takashi
>
>
> On Tue, Jun 6, 2023 at 4:48 PM Herve Beraud <hberaud at redhat.com> wrote:
>
> Hello,
>
> Indeed, Oslo doesn't have PTG sessions.
>
> Best regards
>
> Le lun. 5 juin 2023 à 10:42, Masahito Muroi <masahito.muroi at linecorp.com>
> a écrit :
>
> Hello Herve,
>
> Thank you for the quick replying.  Let us prepare the spec and submit it.
>
> btw, does olso team have PTG in the up-comming summit? We'd like to get a
> quick feedback of the spec if time is allowed in the PTG. But it looks like
> oslo team won't have PTG there.
>
> best regards,
> Masahito
>
> -----Original Message-----
> *From:* "Herve Beraud"<hberaud at redhat.com>
> *To:* "室井雅仁"<masahito.muroi at linecorp.com>;
> *Cc:* <openstack-discuss at lists.openstack.org>;
> *Sent:* 2023/06/05(月) 17:21 (GMT+09:00)
> *Subject:* Re: [oslo] HTTP base direct RPC oslo.messaging driver
> contribution
>
> Hello Masahito,
>
> Submission to oslo-spec is a good starting point.
>
> Best regards
>
> Le lun. 5 juin 2023 à 10:04, 室井雅仁 <masahito.muroi at linecorp.com> a écrit :
>
> Hi oslo team,
>
> We'd like to contribute HTTP base direct RPC driver to the oslo.messaging
> community. We have developed the HTTP base driver internally. We have been
> using the driver in the production with over 10K hypervisors now.
>
> I checked the IRC meeting log of the oslo team[1], but there is no regluar
> meeting in 2023. Is it okay to submit oslo-spec[2] to propose the driver
> directly, or is there another good place to discuss the feature before
> submitting a spec?
>
> 1. https://meetings.opendev.org/#Oslo_Team_Meeting
> 2. https://opendev.org/openstack/oslo-specs
>
> best regards,
> Masahito
>
>
>
>
> --
> Hervé Beraud
> Senior Software Engineer at Red Hat
> irc: hberaud
> https://github.com/4383/
>
>
>
>
> --
> Hervé Beraud
> Senior Software Engineer at Red Hat
> irc: hberaud
> https://github.com/4383/
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.openstack.org/pipermail/openstack-discuss/attachments/20230629/45e92beb/attachment-0001.htm>