Hello ​@TZoltan - changing the MTU affects the whole network interface as that is where this option is located, so it would affect all traffic. Can I ask why you wish to change this option? What issue are you trying to solve with it?

Hi ​@PhillC . Thanks for your answer. I was interested if it can help in an Atlas environment where the latency between a client and the primary can be big.

Currently we are facing problem when connecting to a Windows client can be very slow (40s), while for example connecting to a Linux can be 10s.

I don’t intend to change it, I was just curious what traffic it affects, so thank you for your answer.

​@PhillC just another question regarding this, would lowering the MTU affect BeyondTrust connection performance in a wrong way? Do you know anything about this topic or should I open a ticket to support?

Hello ​@TZoltan - this is a very situational topic, as generally, the MTU setting does not need to be changed unless your local network or ISP network is specfically set up to something different or is an alternative network style.

Based on what you said before - Windows Vs Linux connection start slowness, I would suspect this issue is actually at the client end. We should start by gathering logs on that remote Jump Client (using our BLog.ini) and capture the client logging to see how it responds to the ‘start jump session’ request from the appliance. We can then review what it did and potenionally why it took so long to connect.

Hi ​@PhillC , actually the site is on low Bandwidth and I assumed that maybe Linux Jump Clients needs less traffic (smaller packet sizes, etc..), also connections through Jumpoints are working well. 

What I can see in these logs, that the start of the logs start almost right away when I click  on the client, and around 10s elapses between this and the “Authenticating” line :

notified: oStateReflector:runnerState:changed] value: ]3] reason: Authenticating...]

And then the it starts to communicate with the Traffic Node it’s assigned to (based on network address prefix) for 2s, and when this is done, it’s connecting to a Traffic Node that should not handle the connection. ← that’s what is more strange for me

So basically that’s why I thought lowering MTU could help, because of big latency in communication with the Primary.

We have computers that are servers with big amount of resources (no performance issue) and can produce the same.

Thanks for your help btw.