Over the weekend we have deployed in production attached shema.

All user connected CRS (6.36) are new in network but CCR (6.36) and CRS (6.35.2) connected to servers were deployed already for almost a year and don´t seems to be any problems till now. All have configured MTU 9180.

Users are now experiencing open connection closed to SQL servers (like Server gone away) but I cannot established any pattern why is that happening and when.

Question 1: are there any know problems with 6.36 in CRS?

Question 2: Any tips where to start debugging? trunk problem? MTU problem?

Do you do any kind of NAT or something that requires Connection Tracking?

If yes, check the timeout values at /ip firewall connection tracking

I had similar problems with SSH sessions timing out after being idle for too long.
I increased the tcp-established-timeout setting in connection tracking and I no longer have any disconnections.

By the way, I use 6.36 on dozens of devices for over a month and I haven't had a single problem so far.

I don't remember if the issue I described started on 6.36rc or before. But I also had the connection tracking settings changed some time in the past. I think the default values don't cause any problems like the one I described.

still debuging.....

It is pain in the ass to debug since I cannot establish any pattern when its happen.

there is no NAT in between, just normal LAN.

At the moment looks like some MTU related problems.

I´ll be back when i know more to share details.

SOLVED: bad cables had disturbed the part of network

PS: I still need to figure out how to isolate ports on MT better so they don't disturb the whole network