about a week ago it started... we have huge networking problems
some servers were not reachable (at least it seems... they seem to lose about 99.9% of the packets), some very slow (about 0.5 MBit over the 1 or 10 GBit interfaces)
we had no idea what to do, that's why we did shut down everything... every switch, every server -> power out (drastic, everyone was excited about this, but we didn't have any other ideas anymore)
we did then build a test network by removing everything but 1 server and 1 switch (connected to another switch fo which we think it's ok, including some pcs)
so, we now have -> 1 little network with some pcs connected to 1 switch... from there 1 uplink to our 10 GBit switch and connected to this switch 1 server (2019, hyper-v host) with 2 network adapters... both are connected to 1 port of this switch into a different vlan
on this machine we run 5 virtual computers, one is a 2019 server with an router installed (RRAS) ... this is the one we're testing
we did run a LAN benchmark with a tool we wrote... this tool allows to send or receive data...
if we send, the server first receives and confirms this selection from the client (-> gives us at least 1 packet in and out), then receives a lot of data and creates a report... this report is then sent back to the client
if we receive, the server first receives and confirms this selection from the client (-> also 1 packet in and out), then the server starts to send a lot of data and at the end informs about the end... the client can create the report then itself
... if we now send from the client (x.x.64.10) to the server (x.x.64.1) -> full speed
... same for receiving from the server (x.x.64.10) to (x.x.64.1) -> full speed
if we send to the inner interface ("after" the router) -> (x.x.64.10) to (x.x.0.1) (via x.x.64.1 which is the next hop in this route), we see almost the full speed (the router makes it a bit slower... ok, fine... can understant that)
and now the interesting thing
if we receive from this inner interface ... we have almost no speed... instead of the 850 to 960 MBits we see, we now have only less than 1 MBit ... but not 0 ... and those initializing packet always work... same for pings... no problem... < 1 ms
I did analyze the traffic with the message analyzer... it seems, that we have a lot (really a lot) of re-sends ... in this case... and it takes about 60 seconds to deliver 1 packet ... and things like that... but, it's not impossible
... but wait! that't not all... let's make it a bit more complicated :-) ... now, we used one of those other virtual machines on this server which are virtually connected to the x.x.0.x network and also one physical computer connected to this network... and we benchmarked those... here we see low traffic values (when using the receive mode), but... not that low... more in the range of 300 to sometimes 600 MBits...
And also file transfers via this router now works with 30 MB/s (sometimes) or up to 60 MB/s (sometimes) ... but no longer with 120 MB/s what we had 2 weeks ago.
... any ideas what I can do? I have tested this with multiple switches, multiple network adapters... but, values are jumping around
Rudolf
... oh, by the way... Intel x722 ... :-) is this the problem? ... why only after 3 months of operation?