717073 - sync: reg -> 1.0-3, sreg -> 1.0-13, core -> 2.7-3

Status: VERIFIED FIXED

(Cloud Services :: Operations: Deployment Requests - DEPRECATED, task)

Description

[Toby Elliott [:telliott]]

Train to keep track of the reg and sreg update. 

On both reg and sreg, server-core needs to be updated to rpm-2.7-2. The auth backend should use services.user instead of services.auth. See accountportal for an already switched server.

Reg server needs the server-reg rpm-1.0-1

Sreg rpm remains unchanged.

-----------------------------

QA testing:

The main functional change is to alter how cef events are reported to the backend, but it's been sufficiently long that many pending patches with improvements have been sitting in queue, and the auth lib is switching to user. So, general user creation/deletion, password changes, as well as a few requests with bad passwords and such so that we can check that cef is picking them up correctly. Many of these features have been tested with account-portal, which uses the same core lib.

Comment 1

[:Atoll]

reg:
r26891 2f7ed17 Merged accountportal [auth] changes into reg [auth], copy [reset_codes] from ap, remove deprecated options
python26-services 1.7-2 -> 2.7-2
python26-syncreg 0.5-1 -> 1.0-1

sreg:
r26892 d19f812 Remove deprecated options
python26-services 2.6-5 -> 2.7-2

Gunicorn is up and running on web{1,2}.{reg,sreg}.scl2.stage.svc

Comment 2

[:Atoll]

Ops stage smoke test complete:

created new account with password #1, confirmed /meta/global created in sync-log, changed password to #2 in Firefox preferences, used new password #2 to log into account portal, cleared my sync data, changed password to #1, logged out, forgot password, clicked link in email from [e-mail], changed password to #3, hit Sync Now in Firefox, entered password #3 into resulting credentials dialog, worked.

Assigning to :jbonacci for QA, updated whiteboard.

Comment 3

[:Atoll]

CEF results of my smoke test should be spread across these four servers, username ilxr4syrlw5sryshmqtoz2odvh5mx655

web1.reg.scl2.stage.svc.m.c
web2.reg.scl2.stage.svc.m.c
web1.sreg.scl2.stage.svc.m.c
web2.sreg.scl2.stage.svc.m.c

Comment 4

[James Bonacci [:jbonacci]]

Services QA has picked this up for testing in Stage.

Comment 5

[James Bonacci [:jbonacci]]

I have completed my part of testing this. Final QA sign-off scheduled for 3pm PDT (or sooner).

Details:
I ran through a battery of smoke tests using several current Stage accounts, one re-created account, and one new account.

1. Standard sync across several devices: desktop and mobile
2. Password change from FF, verified across devices and in AP
3. Sync key change from FF, verified across devices and in AP
4. Password change in AP, verified across devices
5. Forgotten password in AP, verified across devices
6. Clear out sync data in AP, verified across devices
7. Account deletion from AP, verified across devices
8. Account re-creation through FF, verified across devices and in AP
9. New account creation and use in FF, AP

Comment 6

[John Morrison [:jrgm]]

I ran into a couple of crashes/dinosaurs while using https://stage-account.services.mozilla.com/

These were c7bd69697c3c13e97373ffb8bb55df0f and 101d5bc743d1487b6e8dd0895eb31a28
both while going through some forgot password sequences.

These were apparently due to stale mysql connections in the backend.
Errors seem to have stopped (since presumably the stale connections have been 
replaced).

Comment 7

[John Morrison [:jrgm]]

I ran through similar sequences to jbonacci above with multiple desktop instances
of FF9, Aurora, Nightly, and aside from the non-blocker errors with stale 
connections in the stage environment, it checks out ok. 

So I guess QA signs off on this work. Thanks.

Comment 8

[:Atoll]

Accepting QA results, deploying next week.

Comment 9

[:Atoll]

https://intranet.mozilla.org/Services/Ops/ChangeWindow_20120124

Comment 10

[Pete Fritchman [:petef]]

this is failing loadtest. all of the webhead->mysql connections are going to sync1.db.

Comment 11

[:Atoll]

(In reply to Pete Fritchman [:petef] from comment #10)
> this is failing loadtest. all of the webhead->mysql connections are going to
> sync1.db.

Bumping to server-reg 1.0-2 to pick up the fix for bug 721558 and deploying to staging to see if that resolves the issue.

Comment 12

[Toby Elliott [:telliott]]

At this point we should just go straight to 2.7-3 and pick up all those fixes, too.

Comment 13

[:Atoll]

Updated topic

Current plan:

reg 0.5-1 -> 1.0-2, core 2.4-3 -> 2.7-3
sreg 1.0-11 -> 1.0-11, core 2.6-6 -> 2.7-3

Comment 14

[Toby Elliott [:telliott]]

sreg 1.0-12 has the new build process changes (and nothing else), so it may be easier to build that one.

Comment 15

[:Atoll]

(In reply to Toby Elliott [:telliott] from comment #14)
> sreg 1.0-12 has the new build process changes (and nothing else), so it may
> be easier to build that one.

reg 0.5-1 -> 1.0-2, core 2.4-3 -> 2.7-3
sreg 1.0-11 -> 1.0-12, core 2.6-6 -> 2.7-3

Comment 16

[:Atoll]

(In reply to Richard Soderberg [:atoll] from comment #15)
> reg 0.5-1 -> 1.0-2, core 2.4-3 -> 2.7-3
> sreg 1.0-11 -> 1.0-12, core 2.6-6 -> 2.7-3

reg 1.0-2 builds rpm 1.0-1, sreg 1.0-12 builds rpm 1.0-11. so, with that fixed, we'll be at:

reg -> 1.0-3
sreg -> 1.0-13

can you fix/verify/tag?

Comment 17

[:Atoll]

I am no longer able to build, filed bug 750824 to get that issue resolved.

Comment 18

[:Atoll]

reg 1.0-3, core 2.7-3
sreg 1.0-13, core 2.7-3

Appears to be working, deployed to staging. Please test all account portal functions so that sreg is fully exercised.

Comment 19

[James Bonacci [:jbonacci]]

QA testing is in progress.

Comment 20

[James Bonacci [:jbonacci]]

QA has completed testing in Stage for this deployment.
Same testing as stated previously plus some extra work around registration, passwords, etc.

Comment 21

[:Atoll]

Deployed to SCL2, PHX1.  https://intranet.mozilla.org/Services/Ops/ChangeWindow_20120509

Comment 22

[James Bonacci [:jbonacci]]

Very thoroughly verified by QA with extensive testing for sreg/reg - new and used accounts.