Sync against the Stage TokenServer is failing due to unsupported /1.1 compatibility route

VERIFIED FIXED

Status

--
major
VERIFIED FIXED
5 years ago
5 years ago

People

(Reporter: jbonacci, Unassigned)

Tracking

Firefox Tracking Flags

(Not tracked)

Details

(Whiteboard: [qa+])

(Reporter)

Description

5 years ago
Running today's Nightly build (which as rolled to Fx30)

Fresh profile with new account using an email account not currently in use in Prod env.
This was working for awhile, but it is now failing for me. 
The symptoms are shown as a yellow bar error:
"Sync encountered an error while syncing: Incorrect account name or password. Sync will automatically retry this action."

I am seeing the following in the sync logs:
GET fail 404 https://sync-3-us-east-1.stage.mozaws.net/1.1/25559/info/collections

Note the "1.1" in the middle there. It should be "1.5" as in
https://sync-3-us-east-1.stage.mozaws.net/1.5/25559/info/collections

This came up in a discussion on the "sync-dev" list in January:
https://mail.mozilla.org/pipermail/sync-dev/2014-January/000800.html
(Reporter)

Updated

5 years ago
Whiteboard: [qa+]
The tokenserver db "services" table was using {node}/1.1/{uid} as its node endpoint pattern.  I updated the column to the correct {node}/1.5/{uid} pattern, Bob did some magic to resatart the tokeserver webhead processes, and this seems to be fixed now.
(Reporter)

Comment 2

5 years ago
Good enough for me.
I got sync to happen correctly (at least for one profile/account/email/device).
Status: NEW → RESOLVED
Last Resolved: 5 years ago
Resolution: --- → FIXED
(Reporter)

Updated

5 years ago
Status: RESOLVED → VERIFIED
(Reporter)

Comment 3

5 years ago
Via bobm:

1. The services table contains records that list the format of URL requests, which happen to contain the version string for the protocol.
The original nodes we rolled out were version 1.0 only, and we put in a hack to make them look like 1.5 nodes.
That information was lingering in the db.
rfkelly removed it, and then I restarted the individual token server apps.
2. the secret restart sauce for circus: circusctl --timeout 30 restart tokenserver
On the individual nodes.
There are a lot of uwsgi workers, and it was taking circus longer than the default timeout period to process a restart.
You need to log in before you can comment on or make changes to this bug.