Network issue, streaming stalls RTL8112L Gigabit LAN


Recommended Posts

2 days ago I had some spare time and decided to upgrade my main server to v6 (6.1.6)

I had no problems with 5.0.6 so far.

 

All went fine until I streamed the first movie to my Popcon C200.

I was facing random pausing in playback (about 10-20 seconds each time).

 

As I have a second server as a backup, I powered it up and all streaming went smooth with this one.

Both servers are running 6.1.6

 

The only difference is the NIC.

No hardware changes were made.

 

The main server has an ASUS M4A785TD-V EVO mainboard with RTL8112L Gigabit LAN

The backup server is a Supermicro - X7SBA with Intel 82573V Gigabit LAN (actually it has 2 Intel NICs)

 

What can I do besides reverting to v5?

Link to comment
  • 1 month later...

Coming back to close this as "temporarily solved".

 

I had good mail support from JonP and we tried to solve the issue.

In addition to the stalls during streaming, I also noted a breakdown in transfer speeds during copy.

Copying to the empty cache drive used to be steady @ ~100 MB/s in v5.

In v6 it is fluctuating between 40 - 60 MB/s (even after upgrading the single core CPU to a dual core).

 

The solution is not yet possible on unRAID side.

JonP mentioned some samba tweaks to come with 6.2 - we will see when it arrives.

 

Until then, I have installed an additional Intel NIC and it seems to work.

 

Drawback and workaround:

My server is not running 24/7 and we use to "WoL" it when we need it.

While the media player is powering up, the server is ready too.

Unfortunately, when using the Intel card, which is plugged in the PCIe x1 slot, there is no WoL working.

This may be a card specific setting, but I didn't dig deeper into it.

Found a post somewhere that said there is a tool that can set cards ROM settings...maybe WoL can be configured

to work then?

But I tried the bonding mode in network settings and it seems to work nicely, so I gave up with the NIC settings.

The setting in unRAID is called "balance - tlb (5)". (tlb=transmit load balancing)

Setting "balance - alb (6)" works also. (alb=adaptive load balancing)

More about linux driver bonding modes: https://en.wikipedia.org/wiki/Link_aggregation#Driver_modes

 

I'm noticing several lost packets on the dashboard error list of eth0 (this is the Realtek NIC), but streaming and copying is nice.

 

EDIT:

added info about the bonding setting.

Link to comment

But I tried the pairing/bonding mode in Network settings and it seems to work nicely, so I gave up with the NIC settings.

 

May I ask what pairing/bonding settings you used and did you use a consumer grade switch?

 

I'm noticing several lost packets on the dashboard error list of one NIC (probably the Realtek and the reason of all issues), but streaming and copying is nice.

 

I've only ever experienced dropped packets with Intel NICs - never with Realtek or Broadcom ones.

 

Link to comment

May I ask what pairing/bonding settings you used and did you use a consumer grade switch?

Sure, but I have to check to give you the correct answer.

It's a more or less cryptic setting from the drop-down list, so I can't remember. I will check when I'm  home tonight.

The switch is nothing special, consumer grade - yes.

 

I've only ever experienced dropped packets with Intel NICs - never with Realtek or Broadcom ones.

Could also be the case. I was just guessing, since I had problems with the Realtek.

I didn't dig to find out which one it is. The dashboard doesn't display the

NIC names. Maybe the diagnosis export reports the true assignment. I will check that also.

 

Link to comment

Edited the OP to add anwers to your questions.

It's load balancing - tlb that I had set.

Now I'm trying alb - out of interest. This is not only transmit load balancing but also receive load balancing.

 

With regard to the lost packets, I see it is the eth0 interface on the receiving side.

Out of the syslog I can read:

eth0: RTL8168d/8111d
eth1: Intel(R) PRO/1000 Network Connection

Link to comment

Interesting, thank you. The bonding issue is indeed complicated and confusing. I thought that only "active-backup (1)" was supported by regular, un-managed switches and everything else caused an ARP broadcast storm unless spanning tree protocol was enabled. I need to read more and experiment more.

Link to comment

Had a similar issue with my Realtelk-NIC.

After waking up from sleep, the NIC was not there and the server not reachable over LAN - strange, but the WOL was working, but all LEDs on the NIC were OFF.

After a reboot it worked fine till the next sleep/wakeup.

I have this problem since uptated from unRAID v5 to v6.

I posted this problem in this forum here, but the developer never answered about this problem.

My OnBoard NIC is a RTL8111E

Then i did a Temp-Install of Windows 7 to investigate this problem but it never happend under Windows.

So it must be "definitely" an unRAID v6 Problem.

But because the Devs never answered my Question, i think, they know this problem and maybe working on it.

Link to comment

Interesting, thank you. The bonding issue is indeed complicated and confusing. I thought that only "active-backup (1)" was supported by regular, un-managed switches and everything else caused an ARP broadcast storm unless spanning tree protocol was enabled. I need to read more and experiment more.

Only mode=4 (802.3ad) requires smart switch with lacp support, all other work with any switch.

Link to comment

Only mode=4 (802.3ad) requires smart switch with lacp support, all other work with any switch.

 

Thanks for that, johnnie. When I turn the Help on in Network Settings it says

 

Mode 1 (active-backup) is the recommended default. Other modes may require switch support.

 

(my italics). I'll do some more reading and then experimenting.

 

Link to comment

I posted this problem in this forum here, but the developer never answered about this problem.

 

As you can see, I also posted here and nobody answered.

I guess you can't expect them to read every topic on the forum.

You have to send LT a support email.

[email protected]

 

I think you are right - thanks for your hint  ;)

 

I  do believe the issue being discussed is S3 sleep.  If this is the case, the issue is well hidden when you look at the topic of this thread.  Another point is that S3 sleep is not directly supported by LimeTech.  This feature is actually implemented by a Plugin.  Over the years that have been many, many problems with Sleep and a lot of it (particularly in the early days) stemmed from hardware issues.  Later hardware designs and BIOS updates apparently have eliminated most of these.

 

As well as appealing to LimeTech, the person(s) with issues should post up a new thread with a Subject that reflects the issue and see if any of the forum members might have some insight into how to fix the issue(s).

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.