push apps failing in Android until you disable IPv6

[JORDI PALET MARTINEZ]

Hi Phil,

Not sure if you have seen the previous message with the rdisc6. Your network may be not having a “broken” CPE.

The problem has been detected with several combinations of different CPEs and Android phones. At the moment I’ve got only rdisc6 from one CPE, but I’m sure others have the same problem (announcing a default route with only link-local), that’s why te problem is sorted out disabling IPv6 in the CPE.

It happens immediately the phone is “sleep”, not after a certain time, etc.

I’m sure is not an IPv6 on the upstream network, because there is no IPv6 there :-(

Saludos,
Jordi








-----Mensaje original-----
De: <ipv6-ops-bounces+jordi.palet=consulintel.es at lists.cluenet.de> en nombre de Phil Mayers <p.mayers at imperial.ac.uk>
Responder a: <p.mayers at imperial.ac.uk>
Fecha: martes, 10 de mayo de 2016, 13:58
Para: <ipv6-ops at lists.cluenet.de>
Asunto: Re: push apps failing in Android until you disable IPv6

>On 10/05/16 09:58, JORDI PALET MARTINEZ wrote:
>
>> Right, but how this is affecting IPv4 push notifications ?
>
>A lot of stuff goes on to make this work. I'm not an expert on the topic 
>of mobile powersaving, but at minimum:
>
>  * The wireless endpints have to be doing WMM (I think) to queue the 
>packets to the endpoint while the client wifi stack is not powered up.
>
>  * The wifi stack on the client endpoint has to be waking periodically 
>and receiving the queued packets.
>
>  * The wifi stack also has to filter the packets to get only the ones 
>which should wake the CPU, and then wake the CPU and deliver them.
>
>  * There is also the Android-specifc Doze state, which suppresses 
>app/network wakes entirely for progressively longer-spaced check periods:
>
>https://developer.android.com/training/monitoring-device-state/doze-standby.html
>
>This is all network, OS, firmware and hardware-revision dependent, and 
>will vary radically from device model to device model (the variety of 
>android devices being a curse as well as a blessing from this PoV ;o)
>
>I've seen problems with this before on an entirely v4-only network with 
>a very early release of Android on Nexus 4 - my debugging showed the 
>problem was likely with the wake filter, and the next release of Android 
>resolved it.
>
>It's theoretically possible that the issue is on the upstream network, 
>with a failure to do WMM in the presence of IPv6. Or perhaps the 
>presence of a certain v6 config causes the device to fail to process the 
>network traffic at the client end.
>
>Or it's an instance of the doze state, which is behaving differently on 
>an IPv6 network for some reason.
>
>FWIW, my Nexus 5 has no such problems on both v4-only and dualstack 
>networks, so I think you're likely seeing some combo of device- and 
>network-specific behaviour.
>
>When customers report these problems, how long has the device been 
>asleep? Is it possible it's dozing?
>
>IMO you'll need to get hands on a device and network combo that 
>demonstrates the problem to get anywhere; there's just not enough info 
>here to determine the cause, only to guess about the range of possible 
>causes.
>
>Regards,
>Phil
>