Re: tv_grab_sd_json: Server closed connection without sending any data back

On 09/12/16 15:57, Robert Kulagowski wrote:
> On Fri, Dec 9, 2016 at 8:51 AM, Jan Ceuleers <[email protected]> wrote:
>> I can't draw any conclusions from the trace since the juicy bits are
>> encrypted but I do see a flurry of retransmissions and the like; perhaps
>> this might be of interest to you?
> 
> Access the load balancer via port 80 instead of 443 so that you can
> see the juicy bits.

Okay, found the problem. (I had to patch the grabber to use http rather
than https but I verified that I could still reproduce the problem after
that change).

There was no reply from the server to the client's POST
/20141201/schedules, not even a TCP ACK. So the client kept
retransmitting these segments every so often until it gave up and reset
the connection.

So clearly a network issue.

I first tried completely disabling the traffic shaper I have on my DSL
interface (to keep bufferbloat in check). This made no difference so I
put it back.

I then disabled ECN (explicit congestion notification) support on the
backend (which is the machine on which the grabber runs), and this
caused the grabber to start working again. So there's a network element
somewhere between me and the AWS servers used by SD that doesn't play
well with ECN.

Strange how this issue only affected the _sd_json grabber and not the
_zz_sdjson one.

Anyway, thanks for sticking with it. Problem solved.

Jan

------------------------------------------------------------------------------
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today.http://sdm.link/xeonphi