L4 metric TCP Flow Stalls


#1

What is the difference between a RTO Out or In (tcp:rto_out and rto_in) and a TCP Flow Stall (tcp:rto_multi_out and rto_multi_in)?

Description in the EH documentation is TCP Flow Stalls: Number of events in which the group was not responsive. Click to break down the number of non-responsive events by each group member in the table at the bottom of the page.

Doesn’t seem to really help on their relationship or what it means; just seems vague.


#2

Hi @davidlreed too many consecutive RTOs leads to TCP flow stall, and you can see the back-to-back RTOs on a flow via the TCP flow stall metric. Does this help? :slight_smile:


#3

So if I drill into a device and I see 229 TCP Flow Stalls (out) that is basically back to back RTOs on that device?

IF that is a correct statement does that mean 229 is really 228 RTOs in that moment, since the Flow stall would be the measurement between back to back RTOs?

I’m just trying to understand the correlation or association between the 2 metrics.

Thank you for your time.


#4

Hey David,

A Flow Stall is actually 3 back-to-back RTOs. You will get a Flow Stall incremented each time we see this event occur. Does that make sense?


#5

Yes that makes sense, thank you for your time.