Re: [Pgbuildfarm-members] Submission failures: 500 read timeout

On Mon, Sep 22, 2014 at 12:21 PM, Stefan Kaltenbrunner
<stefan(at)kaltenbrunner(dot)cc> wrote:
>> data after receiving a SACK packet if the missing data isn't
>> acknowledged during the retransmit timeout.

> hard to say - but that description feels like a common problem going 10
> years backwards when stateful firewalls started doing sequence inspection
> and randomisation but were not yet SACK aware.

That could be an explanation, but AFAICT there is no evidence of that.
The SACK ranges match up to the packets retransmitted after the FIN,
and the eventual HTTP 200 response indicates that the remote did
reassemble the correct stream after receiving those packets. So I see
no reason to suspect middleboxes.

> It might be a long stretch but [...] _AND_
> there is a device on either side that has a slightly broken stateful
> inspection firewall

There's just a Juniper SRX100B doing NAT on my end, which isn't very
old and shouldn't have any paranoia settings enabled.

Regards,
Marti