Hi, I've been having some connection issues lately with my unRAID 6.1.8 server. Usualy I get notified of the problem because a backup has failed or the movie that I'm streaming from plex has stopped. The server simply loses it's connection with everything else in the network. All the other devices, wired or wireless, are still working fine at full speed.

 

Last time that it happened I logged in the console and the server itself seems to be working properly and it's responsive. However from the server I cannot ping anything (other machines, gateway, google.com) even though it still has its IP address (192.168.1.70). I went behind the computer and the network card lights are on. I went into the syslog and I don't see anything wrong at the time of the problem.

 

Everytime this happens, all I have to do is unplug the network cable and replug it and 2 seconds later everything is working fine again

I've tried changing the cable, changing port on the router. I also tried to change the name of the bridge from br0 to br1 since in this thread it might have made a difference in this thread:

http://lime-technology.com/forum/index.php?topic=45909.0

 

In that thread the person had to change the network card to solve his problem because of driver issue. However I believe the broadcomm chipset is supported in the wiki page of unraid and I would appreciate troubleshooting the issue a little bit befonre going to buy another NIC.

 

Here is my config running unRAID 6.1.8:

Dell Precision T5500

2x Intel Xeon E5620 @ 2.4ghz

12gb of ECC DDR3 ram

Nvidia quadro 600 video card

Ethernet controller: Broadcom Corporation NetXtreme BCM5761 integrated

1 LSI Logic 9211-8I raid card (nothing plugged on it yet)

1 2tb WD Red parity disk

3 2tb WD Green data disk

1 2tb WD Red data disk

 

Dockers:

Plex (limetech/plex:latest)

qbittorrent (mace/qbittorrent:latest)

RDP-Boinc (aptalca/docker-rdp-boinc:latest)

 

VM:

Windows server 2012 essentials (named virtucube) running through a bridge interface br1

 

Yesterday it did it again between 4pm and 1am (I know it's a large interval but I wasent' home). The router log doesn't seem to log anything that would drop the connection:

 

Mar 27 16:59:59 ntp: start NTP update

Mar 27 17:02:14 miniupnpd[11631]: upnp_event_process_notify: connect(192.168.1.160:2869): No route to host

Mar 27 17:02:14 miniupnpd[11631]: upnp_event_process_notify: connect(192.168.1.160:2869): No route to host

Mar 27 17:59:57 ntp: start NTP update

Mar 27 18:34:45 dnsmasq-dhcp[11573]: Ignoring domain CYBWARE.local for DHCP host name VirtuCube

Mar 27 18:59:57 ntp: start NTP update

Mar 27 19:59:59 ntp: start NTP update

Mar 27 21:00:01 ntp: start NTP update

Mar 27 21:59:57 ntp: start NTP update

Mar 27 22:59:57 ntp: start NTP update

Mar 27 23:59:59 ntp: start NTP update

Mar 28 01:00:01 ntp: start NTP update

Mar 28 01:59:57 ntp: start NTP update

 

I've included 2 syslog, the first one (0105.zip) is when the server wasn't responding to the network, the second one (0115.zip) after I unplugged the network cable and replugged it. I hope someone can make sence of all this and help me find the cause.

 

I hope I've included enough information and thank you for your time.

tower-diagnostics-20160328-0105.zip

tower-diagnostics-20160328-0115.zip

When the connection is working, have a look at the lights on the CAT5 socket on your server and the lights connected to the port on the switch at the other end.  Recorded that information.  Then when you notice that you have 'lost' the connection, do the same thing.

 

I will now tell about an experience that I had.  I had a problem with exactly the same symptoms as you are having.  What I found was that I unplugged and plugged AC Power to the switch, normal operation was restored.  I played around with changing cables and that sort of thing.  I finally decided to 'bite-the-bullet' and bought a new switch.  (It was a twenty-four port switch so it was not painless on the pocketbook.  The switch that I purchased was a cheap no-name switch.)  About eighteen months later, same problem.  Again, it was the switch.  I looked and double checked reviews of user experiences and purchased a switch this time with better reviews.  I have not had a problem with this new switch to date....

 

What I think is the issue is with the 'Green' feature that most modern switches (and routers) have.  There is circuitry inside of them that shuts down ports that have nothing connected to them and modulates the signal level so that shorter cables are driven with less power.  I contend that these 'features' sometimes serve to disconnect ports that have long periods of inactivity.  I don't know if this is your problem or not but keep in mind that it may well be...

Thank you for your reply. I actualy tried something like this. My server was originaly plugged into a trendnet 8 port gigabit switch and I've since moved it into my asus ac1900 router directly. Unfortunately the problem persists.

 

Now that I think about it, the problem might have started when I added a VM on the server using the bridge interface.

I could not make it work with the virtbr0, I prefer having my vm server on the same network as everything else.

 

I guess I could try to remove the VM and see if the problem persists. That would be unfortunate since I planned to use unraid to merge my windows server and my storage.

I believe I have fixed my connection loss problem. Here is what I changed.

 

I bought an Intel PRO1000 dual port pci express NIC to replace the onboard broadcom NIC to see if my issue would disappear. Nope, I still lost connection when my VM was passing through the bridge to access the network.

 

I finaly decided to try a PCI passthrough the onboard NIC directly to the VM and using the new dual intel pro1000 (with bonding because why not) for unraid shares and since then my server has been up 24/7 without any issue.

 

Here is the guide to enable pci passthrough for network cards to VMs, don't forget to remove the bridge connection of the VM when adding the pci passthrough in the xml config :

https://lime-technology.com/forum/index.php?topic=39638.0

This may be worth posting a bug report for in the forum here.

 

Worthwhile reading the sticky threads in this forum to ensure you post all the required information.