lagg0 Not Performing

Status
Not open for further replies.

dddza

Dabbler
Joined
Apr 24, 2012
Messages
25
Hello,

Build FreeNAS-8.3.0-RELEASE-p1-x64 (r12825)
Platform Intel(R) Xeon(R) CPU E5645 @ 2.40GHz
Memory 32744MB

Having a performance issue with our lagg0 which was setup accordingly:

lagg0 lagg0 False 10.11.12.150/24 False

The interfaces were not configured before the lagg0 was configured:

LACP was used and the D-LINK DGS-1210 was also configured for these ports to be LACP.

Members:

igb1, igb2, igb3
Options: mtu 9000 up
LAGG Priority on each interface: 0

The machine was also rebooted multiple times after the lagg0 was setup.

The network graphs however look as follows and has not appeared to be bonding

http://i.imgur.com/j78wvDM.png

Any assistance would be much appreciated as we are sitting with massive network congestion.
 

dddza

Dabbler
Joined
Apr 24, 2012
Messages
25
A simple search for "lagg" in the networking section of the forums turns up a LOT of stuff that would have been helpful for you to read:

http://forums.freenas.org/showthread.php?10082-LAG-(LACP)-Created-and-UP-but-not-working

Thank you for your reply. I have done a fair amount of searching and reading up on LAGG and the bonding of LACP. The specific link you gave me indicates that lagg0 is not even reachable. My bonding is perfect from a connection point of view, its handling traffic no problem however, looking at those graphs, surely in a bond they should represent somewhat the same? (ignore igb0 as that's our management).

My concern is that I've done everything to the book on a switch level and FreeNAS regarding bonding using LACP but according to the graphs, and especially when doing something high traffic, only 1 port is ever really maxed out and the others aren't doing much.

Thank you in advance :)
 

dddza

Dabbler
Joined
Apr 24, 2012
Messages
25
Does anybody have any suggestions or any ideas other than "refer to other searches" as I do not appear to have the same issues as others with lagg0. My lagg is apparently working but the traffic is not fully bonded and is only happening over a single interface and it swaps around. The graph shows that other traffic is happening over the other interfaces as part of the same lagg0 so I'm very confused.

I now have two FreeNAS machines up and running. I then transfer a large file from one to the other, and as seen in the attached graphs, bonding is just acting so weird.

Here is one attempt from FreeNAS-1 to FreeNAS-2 over a bonded ethernet on both sides with LACP support on the switch enabled:

http://i.imgur.com/LfnvZgi.png - First Set is Reporting on FreeNAS-1, Second Set is Reporting on FreeNAS-2

Here is a second attempt and again from FreeNAS-1 to FreeNAS-2:

http://i.imgur.com/eHoFgZV.png - First Set is Reporting on FreeNAS-1, Second Set is Reporting on FreeNAS-2

I really need to get bonding to work to improve performance as our entire network is bonded and performance is becoming a problem :(
 

Attachments

  • san2-3-1.jpg
    san2-3-1.jpg
    6.1 KB · Views: 242
  • san2-3-2.jpg
    san2-3-2.jpg
    6.1 KB · Views: 245

jamoses

Cadet
Joined
Sep 22, 2011
Messages
9
dddza,
..and EVERYONE ELSE that is using FreeNAS with LAGG.

I have seen exactly the same problem with LAGG underperformance. I never see faster than 344mbps on large file transfers or more than one interface performing.

I have done this:
1. Established NAS1 and NAS2 machines. Identical hardware and latest FreeNAS.
2. Established an LACP LAGG on NAS1 with em0 and re0 which are both 1Gbps interfaces.
3. Established an LACP LAGG on NAS2 wiht em0 and re0 which are both 1Gbps interfaces.
4. When I type ifconfig, I can see a normal status of 1c<ACTIVE, COLLECTING,DISTRIBUTING> on both em0 and re0
5. I have the switch configured to bond ports for the LAGG interfaces.
6. I KNOW the network bonding for the LAGG is correct. I can even yank lines and watch things switch over and heal when restored when I type ifconfig.

7. My test was to move DVD.iso files via FTP protocol from NAS1 to NAS2. I did it once just to get it in the cache. I did it again, and again. I also tried establishing multiple simultaneous socket connections and simultaneously copied the same file DVD.iso to different target directories from NAS1 to NAS2.

8. In every case my aggregate speed NEVER is faster than 43MBps (344mbps). Also, only one gigethernet port transfer the data at a time. Funny thing is that on one of the NAS servers it's em0 doing the transfer, and the other it's re0.

The motherboard is an ASUS P8P67 EVORev3 with onboard dual 1Gbps ethernet adapters. All drives are 3Tb in a zfs ver28.

Is there anyone that has got this to work?
What is the FASTEST you have seen data transferred?
What protocol and methodology did you use?

Thanks,

-JM
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,681
So, you have two Realtek ethernets in a link aggregation and you wonder why you aren't seeing increased performance from one host to another.

First, you need to check to see how many hamsters are dead. Best to replace that junk with a decent Intel server-grade ethernet if you want to do something high performance.

Then you need to read 802.3ad and figure out that link aggregation does not do what you think it does. It cannot increase the bandwidth for a single conversation. So your single FTP session will not go faster than 1Gbps even if you get out and push.

Link aggregations are only useful if you have one server talking to multiple hosts, preferably over quality ethernet hardware. You are unlikely to fully realize the benefits of link aggregation until you get out to maybe a dozen clients that are moderately active with respect to the server. That's a practical observation rather than a technical one.
 
Status
Not open for further replies.
Top