How to debug connection issues

[rallisf1]

I have MC installed in an Oracle A1 VPS as per official documentation with mongodb and nginx as reverse proxy (I only serve MPS directly from MC).

My problem is that almost all agents momentarily lose connection but they don't go offline according to logs. It is random (not all agents go down at the same time), happens more on linux agents than windows and for that 1-2 seconds they appear grey and all action buttons (desktop/terminal/files) disappear and then come back. If I happen to be connected to agent's desktop/terminal/files the session might freeze for a second and continue like it didn't happen but it won't disconnect me.

Anyone having a similar issue? How should I debug this?

[Ylianst]

A few things here. On Windows, you can run the agent like this:

meshagent.exe run

This will run the agent from the command line and you will see the agent connecting, retrying, etc. This can help with debug. Also, the agent will connect to the server using a long standing websocket connection. Some firewalls, HTTP proxies, load balances, reverse-proxies, etc do not like idle connections and will automatically close them. To fix this, try one of these lines in the settings section of the config.json:

"agentping": 30,            <-- Ping the agent every 30 seconds and have the agent reply with a pong response.
"agentpong": 30,            <-- Pong the agent every 30 seconds, agent will not respond.
"browserping": 30,          <-- Ping the browser every 30 seconds and have the agent reply with a pong response.
"browserpong": 30           <-- Pong the browser every 30 seconds, agent will not respond.

The goal with these lines is to fix your problem while generating the least amount of traffic as possible. So, if you take a stopwatch and see agents reconnecting every 3 minutes, put a "agentpong" at the 2 minute 50 second mark and see if it fixes it. Only use "agentping" if required. Do not use a "ping" and a "pong" to the same target a the same time, it's just a waste.

Let me know if that works,
Ylian

[rallisf1]

I always wondered what these ping-pong settings were about, will try it through the weekend and post an update. Thanks!

[rallisf1]

I tried "agentpong": 30, first, as it generates the least traffic and although the random "disconnects" stopped, I got a big delay when starting any service (desktop/terminal/files). Replaced it with "agentping": 30, and seems to work as expected.