High number dropped rx packets

[bkasten]

FWIW I fired up Wireshark on my PC and set it to monitor retransmissions;  then copied a large file to/from each of my UnRAID servers.  No retrains -- so it's "working perfectly" from the PC side.

 

But it'd still be interesting to know whether the same is true on the UnRAID end ... so a quick tutorial would be very much appreciated.    From a bit of Googling/reading it seems it's just a matter of using iperf and netstat; but I have no idea exactly what commands they need to do this.

 

I discovered this dropped packet thing by accident on my server. I was actually fixing a failed on-board NIC in my computer.

 

Not sure this is a tutorial, but what I did was:

 

Login to the unRAID and issue this command.

 

watch ifconfig

 

It should display something like this:

 

eth0      Link encap:Ethernet  HWaddr 00:30:18:ad:e3:b6 

          inet addr:10.0.1.10  Bcast:10.0.1.255  Mask:255.255.255.0

          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1

          RX packets:2178296621 errors:0 dropped:4301775 overruns:0 frame:0

          TX packets:2070126428 errors:0 dropped:0 overruns:0 carrier:0

          collisions:0 txqueuelen:1000

          RX bytes:2310594674 (2.1 GiB)  TX bytes:3708961407 (3.4 GiB)

          Interrupt:43

 

The advantage of using "watch" is it updates every two seconds.

 

Then sit and watch the dropped packets. Mine increment about 2 every two seconds or so. To cancel hit Control-c.

 

I suspect that more folks than peterb and I have this issue. We just happened to see it.

 

I am not even sure this is an "issue". My server works. It works about as good as it ever has (which is very good for three years). I just saw this when I was fixing an unrelated Ethernet issue.

 

It has been suggested that this may be a "feature" of the RC candidates. Maybe; maybe not. I don't remember ever seeing this on the 4.7 release. I might just be a little OCD on this.

 

Bruce

 

[PeterB]

As I was typing my previous reply, I noticed that my current dropped packet count was only 106 - much lower than I've seen since I noticed this problem.  So, I did a file copy through wireless interface and was astounded to find that the dropped packet count didn't increase.

 

I wondered what had changed - I then remembered that my 'temporary' switch was still in place with only two connections - my unRAID server and my main switch.  I moved the unRAID connection back to my main switch and, even without a file copy, other network activity was causing the dropped packet count to shoot up - several thousand in less than a minute.

 

I moved the unRAID connection back to the 'temporary' switch and the dropped packet count stopped incrementing again.

 

My rationale for this is that there are now two switches in the link, both of which are buffering packets.

 

So, it seems that this problem can be ameliorated, if not fixed, by increasing some buffer somewhere.  More investigation and testing seems in order!

 

To bkasten - it would seem that any device connected via one of your subsidiary switches is not going to contribute (significantly?) to the dropped packets.

[bkasten]

PeterB

 

I wanted to report back to you about the new Intel NIC. I finally got it installed, and all is working fine.

 

Copies are faster through-out the network to from the unRAID. I am thrilled!

 

And the dropped packet count? Still the SAME!!  :'(

 

I turned off each computer, one by one in the house. No difference.

 

Not even sure what to think at this point.

 

Bruce

[garycase]

Is the dropped packet issue restricted to transfers FROM 1000Mb to 100Mb devices?    i.e. is everything okay when going between 1000Mb devices and from 100Mb to 1000Mb ?

 

And if that is the case, do you get the same issue if the devices are directly connected (e.g. no intervening switch) ??

 

[PeterB]

Is the dropped packet issue restricted to transfers FROM 1000Mb to 100Mb devices?

 

Yes but, strangely, the dropped packets occur on the reverse link - ie, tx file from unRAID server (1Gb) to a 100Mb client and the dropped packets occur on rx at the unRAID server.

 

i.e. is everything okay when going between 1000Mb devices and from 100Mb to 1000Mb ?

 

Yes.

 

And if that is the case, do you get the same issue if the devices are directly connected (e.g. no intervening switch) ??

 

Yes.  But having two intervening switches effects a drastic reduction in dropped packets.

[garycase]

Really strange -- but the real performance question is Does it have any impact on performance?

 

i.e. what kind of transfer rate are you getting when sending files from UnRAID to a computer with a 100Mb connection?    If it's ~ 11MB, then clearly the packet issue doesn't matter vis-à-vis performance.