How to diagnose network problem?

[raulik]

Hi,

could someone help me with diagnosing network in unraid? I have small linux websocket client, which connect to somewhere for days.. When I build it for win64

(pc on the same network as inraid), it run ok for days. When I run it on my server in housing, it runs ok for days - both linux/win64 binary. But..

When I run it on my unraid server - in docker, in VM, or on unraid directly, websocket is closed as Connection Closed gracefully random times - mins or hours, but, web socket is closed after mins of nothing received (In the same time, on Windows it didn't disconnect). This client have only one websocket connection, only from unraid to internet. I looked in dmesg, there is no message in this time. I tried to run docker network as bridge, br0, host... Also upgraded to actual rc, but nothing changed. As I have onboard realtek, and Asus 10gbe network, I tried to use 10Gbe as primary and switch off realtek, but also nothing changed.

Something to test?

Thanks

[ChatNoir]

Your diagnostics could provide more informations. 

[raulik]

Hi,

after lot of research, lof of wiresharking I see, that when websocket server send (from time to time) burst with more packet/s there tcp receive window free bytes are going to zero. I dont understand, why this receive buffer is not increased fast, when the size is not as big as in /proc/sys/net/ipv4/tvp_(r/w/)mem. I dont understand linux, not a guru, but, simple comparing windows/linux/unraid(docker,vm,nativ) shows me something bad with unraid tcp stack, or any default configuration there.

 

I also go through diagnostics files, but nothing found. I thing, that this is deep cfg options somewhere, when simple windows or ubuntu (nativ/vm) get this good for days, unraid for minutes