CTI TAC Meeting Notes 2024-06-26

CTI TAC Meeting Notes 2024-06-26

[Carlos O'Donell]

​CTI TAC Meeting Notes 2024-06-26

Present:
 * Carlos O'Donell
 * Konstantin Ryabitsev (LF IT)
 * David Edelsohn
 * Siddhesh Poyarekar
 * Bennet Pursell (OpenSSF)
 
Regrets:
 * Nick Clifton
 
Agenda:
 * Schedule going forward.
   * Done. CTI TAC reviewed SOW and provided feedback.
   * Done. CTI TAC meeting review of remaining SOW items resolved.
   * LF IT SOW sent to OpenSSF (Bennet)
   * AI: Konstantin to send Carlos and Bennet final SOW to share with community (2024-06-26)
   * In progress: 2024-06-27: Sharing early draft plan with the community.
 * Next steps revised:
  * Late June - Share early draft plan with the glibc community.
  * Late July - Agreed migration plan.
  * Late July / Early August - CTI GB would read out the cost and approve in a public meeting.
  * August 15th - OpenSSF GB meeting
  
Notes:
 * Konstantin: Working with OpenSSF to narrow down the SOW.
 * Carlos: What about ROSI pricing?
 * Konstantin: OpenShift licenses in particular.
 * Carlos: What I want is a redacted SOW that I can share with the community?
 * Carlos: We can redact the confidential pieces, just the services is OK.
 * Konstantin: Bennett?
 * Bennett: YEs.
 * Konstantin: Provide SOW? Sure. Is this binding?
 * Carlos: No.
 * David: No.
 * Siddhesh: No. Doesn't have to be binding. Intent is to share a plan.
 * Konstantin: Does the SOW include timing?
 * Carlos: Scope is OK. High level schedule is possible?
 * Bennett: High level of what we want?
 * Konstantin: It depends on availability of hiring?
 * Carlos: We can make a more comfortable statement like "working with LF IT as quickly as we can"
 * Konstantin: We can say we will be starting in Q3?
 * Carlos: Yes the blocking windows are July and January
 * Konstantin: Other things can be done piecemeal?
 * Carlos: Yes.
 * Konstantin: Table of first 3 months has costs of initial deployment.
 * Carlos: Community will ask for the rough number.
 * Konstantin: Estimated annual $276K USD, projected hosting cost and support from LF IT.
 * Konstantin: The LF IT services are extensive and built in to that cost.
 * Konstantin: Is there an SLA requirement?
 * Carlos: No SLA, commercially reasonable effort.
 * David: No, now is not the time to ask for something we haven't required over the last 30 years.
 * Siddhesh: No SLAs.
 * Carlos: We recently lost git push issues due to AdaCore hooks. Routing issues to the colo.
 * David: The SLA is within the LF, with the OpenSSF funding, and the OpenSSF saying this is important from a security perspective.
 * Bennett: We have redundancy and is that something we want to publish?
 * Konstantin: We do mention that already. We can also put in the design of the platform. Automatic failover.
 * Carlos: Yes please.
 * Konstantin: OK, I'll put it into the part.
 * Carlos: Goal to post to the mailing list?
 * Konstantin: Post to you and Bennett and OK it?
 * Carlos: Do you expect me to redact it?
 * Konstantin: Yes, we would hash out what needs to go to the community.
 * Carlos: Yes, I can do that.
 * Konstantin: That unblocks the SOW, and for OpenSSF we will provide more details to sign off on it.
 * Carlos: This week? It's the TACs purppose to review scope? I'm just going to work to redact the confidential data.
 * David: No other requests outstanding. Add text? Send to Carlos?
 * Carlos: Look forward to the email.

-- 
Cheers,
Carlos.

Re: CTI TAC Meeting Notes 2024-06-26

[Frank Ch. Eigler]

Hi -

> [...] Routing issues to the colo.
>  * David: The SLA is within the LF, with the OpenSSF funding, and the OpenSSF saying this is important from a security perspective.
>  * Bennett: We have redundancy and is that something we want to publish?
>  * Konstantin: We do mention that already. We can also put in the design of the platform. Automatic failover.
>  * Carlos: Yes please.
>  * Konstantin: OK, I'll put it into the part.
> [...]

Given that the routing issues mentioned were upstream of the
colocation facility itself, effective improvements in failover would
have to work in a cross-colo way.  Is the intent to have
failover-capable servers in multiple colo facilities?

- FChE

Re: CTI TAC Meeting Notes 2024-06-26

[Carlos O'Donell]

On 6/27/24 10:25 AM, Frank Ch. Eigler wrote:
> Hi -
> 
>> [...] Routing issues to the colo.
>>  * David: The SLA is within the LF, with the OpenSSF funding, and the OpenSSF saying this is important from a security perspective.
>>  * Bennett: We have redundancy and is that something we want to publish?
>>  * Konstantin: We do mention that already. We can also put in the design of the platform. Automatic failover.
>>  * Carlos: Yes please.
>>  * Konstantin: OK, I'll put it into the part.
>> [...]
> 
> Given that the routing issues mentioned were upstream of the
> colocation facility itself, effective improvements in failover would
> have to work in a cross-colo way.  Is the intent to have
> failover-capable servers in multiple colo facilities?

Yes, correct, as noted in the just published SOW:
https://lore.kernel.org/cti-tac/2186f48a-12cd-4630-8944-26b9042e924a@redhat.com/T/#u

-- 
Cheers,
Carlos.

Re: CTI TAC Meeting Notes 2024-06-26

[Frank Ch. Eigler]

Hi -

> > Given that the routing issues mentioned were upstream of the
> > colocation facility itself, effective improvements in failover would
> > have to work in a cross-colo way.  Is the intent to have
> > failover-capable servers in multiple colo facilities?
> 
> Yes, correct, as noted in the just published SOW:
> https://lore.kernel.org/cti-tac/2186f48a-12cd-4630-8944-26b9042e924a@redhat.com/T/#u

Can you point out where that is spelled out?  I see "it will be
deployed within a single public cloud environment", and am missing 
a part which talks about replication of git etc. across multiple
colocation facilities.  (e.g., as far as I'm aware, grokmirror git
replicas are read-only.)

- FChE

Re: CTI TAC Meeting Notes 2024-06-26

[Carlos O'Donell]

On 6/28/24 6:00 PM, Frank Ch. Eigler wrote:
> Hi -
> 
>>> Given that the routing issues mentioned were upstream of the
>>> colocation facility itself, effective improvements in failover would
>>> have to work in a cross-colo way.  Is the intent to have
>>> failover-capable servers in multiple colo facilities?
>>
>> Yes, correct, as noted in the just published SOW:
>> https://lore.kernel.org/cti-tac/2186f48a-12cd-4630-8944-26b9042e924a@redhat.com/T/#u
> 
> Can you point out where that is spelled out?  I see "it will be
> deployed within a single public cloud environment", and am missing 
> a part which talks about replication of git etc. across multiple
> colocation facilities.  (e.g., as far as I'm aware, grokmirror git
> replicas are read-only.)

Sorry, perhaps I misunderstood your question.

Are you asking if we will have authenticated 'git push' capabilities across
multiple co-location facilities?

-- 
Cheers,
Carlos.