Firefox Accounts DevOps next steps for November

Benson Wong bwong at mozilla.com
Mon Nov 4 11:09:51 PST 2013


We have RDS working on a few things, ie: the Location project. 
Not much that I know of other than that. 

If we want multi-region failover, we'll have to figure out what we really expect. If we can't make it happen w/ RDS, we can run MySQL ourselves, or maybe a fork of it with more replication features. 

----- Original Message -----
From: "Chris Karlof" <ckarlof at mozilla.com>
To: "Benson Wong" <bwong at mozilla.com>
Cc: "Gene Wood" <gene at mozilla.com>, dev-fxacct at mozilla.org, "Mozilla Services Operations" <services-ops at mozilla.com>, "Ryan Kelly" <rfkelly at mozilla.com>
Sent: Monday, November 4, 2013 10:56:25 AM
Subject: Re: Firefox Accounts DevOps next steps for November




On Nov 1, 2013, at 4:56 PM, Benson Wong <bwong at mozilla.com> wrote:

> Hi, 
> 
> I think the first step is figuring out what things look like from a high level: 
> https://mana.mozilla.org/wiki/display/SVCOPS/Firefox+Accounts+Server+Layout
> 

I see that you're proposing RDS as our MySQL deployment.

Questions:
1) What's our general experience with RDS?
2) I don't think multi-region operation is a requirement for us at the moment, but it might be in the future. Can we continue to use RDS (which doesn't support multi-region repl), say by replicating to our own MySQL instances in other regions [1], or would we migrate away from RDS?

-chris

[1] https://aws.amazon.com/about-aws/whats-new/2013/09/05/amazon-rds-new-data-migration-capabilities-mysql/



> Then we can create the cloudformations. Let's stick w/ JSON. 
> Then we can craft the puppet stuff, scripts to build RPMs, etc. 
> 
> There are a few goals I would like to do w/ FxA. I'm very flexible on these and willing 
> to let them slide to get the app live (first priority).
> 
> - empower Devs/QA to be able to deploy their own stacks at specific versions of the source.
> - use a svc. ops shared service, heka? statsd? something
> - integrate deeper w/ StackDriver. Perhaps even NewRelic's node.js client (if it's good enough)
> - use Simple Work Flow + web interface for a GUI based deploy. Persona access controlled.
> 
> I'm mostly done w/ the Telemetry guy's ops. More accurately, they not blocked on ops stuff right now so this is my main focus. 
> 
> Let's set up a meeting time next week to go over architecture w/ Ryan? 
> 
> Ben
> 
> 
> ----- Original Message -----
> From: "Gene Wood" <gene at mozilla.com>
> To: "Mozilla Services Operations" <services-ops at mozilla.com>, "Ryan Kelly" <rfkelly at mozilla.com>
> Cc: dev-fxacct at mozilla.org
> Sent: Friday, November 1, 2013 4:21:08 PM
> Subject: Firefox Accounts DevOps next steps for November
> 
> Now that the MySQL connector code has been landed in the API Server for
> Firefox Accounts we (Services Ops) should be clear to assemble the
> provisioning for Firefox Accounts. I've begun work on the
> CloudFormation/Cumulus environment provisioning code.
> 
> Benson, would it make sense for you to start on the masterless Puppet
> provisioning code for the API Server at this point? I suspect that the lion
> share of the provisioning behavior needed is represented in Ryan's bash
> scripts here (
> https://github.com/mozilla/picl-deployment/tree/master/buildscripts ).
> 
> Ryan, can you give some more details on which of those scripts come into
> play on the API server and in what order?
> 
> -Gene
> _______________________________________________
> Dev-fxacct mailing list
> Dev-fxacct at mozilla.org
> https://mail.mozilla.org/listinfo/dev-fxacct




More information about the Dev-fxacct mailing list