C2901 PCAP. What explains this seemingly incomplete stream?

[MicJameson1]

Hello. Please see pcap taken at 2901 device.

2 servers are conducting a stress test at ends of Cisco 2901 routers, and Nexus 9300 switches. The server says there are disconnections at the other server end. 

QUESTION: Why does tcp stream 91 have such a seemingly incomplete stream? What is going on here?

Thank you.

[Joseph W. Doherty · Accepted solution]

What kind of stress test?  One including bandwidth stress testing?

If so, bandwidth stress testing generally causes drops, and impact of drops is also generally adverse to the ends of impacted data streams.  How adverse, by kind and/or degree, depends on multiple variables and how end hosts try to deal with the adverse impact, and BTW, many host apps, if they see too many errors will just "throw in the towel", i.e. give up.

If you're wondering whether I'm using hyperbole, start to research all the various "flavors" of TCP (a dated comparison, there are more later variants) and how they differ, generally how they deal with traffic throughput/congestion.  There's also more when you go beyond TCP.

[Joseph W. Doherty]

BTW, assuming this is in someway related to military networking, there may be a need for "this cannot fail", and if so, I suspect someone can fully identify the root cause and perhaps find a way to preclude it, but if there is not such a need, then perhaps just knowing if you overstress the network can cause issues, is sufficient.

A stressed network, having issues, might only be surprising to those who don't realize such stress issues are not uncommon.  For anyone in the military, with any knowledge of usual machine guns like the old M60 or M2, what happens if you keep firing, not firing in bursts, and not swapping barrels?  I.e. there are limits.  But, it may be possible, if needed, to preclude network stress issues, just as the Vickers machine gun, with a water cooled barrel, has some interesting long continuous firing records.