Closed Bug 1049693 Opened 10 years ago Closed 1 year ago

deploying pushgo may cause signed in FMD devices to lose connection

Categories

(Cloud Services Graveyard :: Server: SimplePush, defect)

defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: edwong, Unassigned, NeedInfo)

Details

We pushed pushgo yesterday and it seems that 2 b2g devices that were already connected were no longer receiving push notifications.  This is hard to repro as you only get one chance against prod.

1. sign in and enable FMD on b2g device
2. restart or push new build of pushgo to prod
3. on desktop browser goto: find.firefox.com

actual: phone doesn't receive push notifications, phone is not found

expected: deploying a new build should not disrupt the ability to find a phone.

work around: toggling fmd enabled button fixes this and re-signing into find.firefox.com.
Push works by holding open a WebSocket connection between a device and the server. If the device loses the connection for any reason (including remote server reset), it should reconnect. What may be happening is that the device is not receiving or handling a socket connection error where the network connection has not also failed. There is no way for a server to be maintained and also be able to preserve this connection.

nikhil, is there potentially a client bug around this? Might require doing local DNS tricks to point to a different mock push server we can bounce up and down.
Flags: needinfo?(nsm.nikhil)
This is all really TCP's fault :)

If the STR was done soon after the redeploy, FMD wouldn't work. The client currently tries to ping the server every 30 minutes, so the earliest it can detect a disconnect (if TCP does not notify earlier) is 30 minutes. Bug 894879 changes the ping interval to a 'reasonable' value for a given mobile network, but that is also upper bound to 30 minutes. I'm not sure what more the client can do about this except to ping more aggressively, but that is a battery life trade-off.
Flags: needinfo?(nsm.nikhil)
:nsm this sounds reasonable. Peter, can we test this on stage upon next deploy?
Flags: needinfo?(pdehaan)

Product obsolete. Closing.

Status: NEW → RESOLVED
Closed: 1 year ago
Resolution: --- → WONTFIX
Product: Cloud Services → Cloud Services Graveyard
You need to log in before you can comment on or make changes to this bug.