Unsolved

This post is more than 5 years old

8 Posts

1238

May 28th, 2016 10:00

powerconnect 8024F and 5524 operating at 100Mbps

We have a situation where it appears that all traffic communicating over our 8024F and 5524 switches is transferring at 100Mbps.

I'm not a networking guy and we outsource our networking to consultants who have not been available to help much, so I'm hoping I can find some guidance here.

Our configuration is as follows:

Two 8024F switches connected to two Equalogic 6010 SANs.
Two 5524F switches connected to the 8024Fs, as well as a Cisco 5520 ASA.
Four Dell R610 servers running VMWare 5.5 these are connected to both the 8024F and the 5524F

The 5524F switches connect to three 100Mb interfaces; one for an EPL to our office from the datacenter, one for internet, and one to backup datacenter.

iSCSI for Dell hosts is directly connected to the 8024F, and the ethernet ports on the Dell servers are connected to the 5524F - ethernet was set up primarly as a backup network.

There are several VLANs; one for iSCSI, one for VMotion, several for segmented data networks, and one for the backup network.  All of the ports on the 5524 are trunked; VLANs are defined on the ASA.

What we're seeing is that throughput to the Equalogics is maxed out to 100Mbps.  I ran a SQL backup this morning and it was only doing 12MB/s which equates to about 96Mbps, and I saw high disk queue length on the server itself and in VMWare.

Monitoring through Solarwinds of the 5524F shows several 1Gbps interfaces all hitting *exactly* 10% utilization.  All of this transmit data; virtually nothing being received.

Oddly, the vSwitch on VMWare for the backup interface (connected to ethernet) shows tons of receive traffic even though the particular host I was monitoring has no guests configured for that network.

Finally, I ran a packet capture on the VMWare host and saw a ton of iSCSI traffic...obviously nothing on that interface accepted the traffic as there is another vSwitch dedicated to iSCSI at 10Gbps.

Our network consultant did suggest that we define all of the 5524 ports for specific VLANs which sounds right.  He looked at the spanning tree configuration and said everything looks good there.

But what I can't understand is why everything seems to be pegged at 100Mbps; none of the 100Mbps show any significant traffic; all of the traffic seems to be over the 1Gbps interfaces.

Thanks for any help you folks can provide!

Eric

Community Manager

 • 

9.6K Posts

 • 

42.8K Points

May 30th, 2016 08:00

Hi,

Are the link lights on the switch green or amber? Page 40

Are these long cabling runs? If you bypass the seitches and connect a laptop right to the devices do they operate at gigabit speeds?

8 Posts

May 30th, 2016 09:00

Hi Josh,

Thanks for your reply.  These are not long cable runs by any stretch - all of the equipment is mounted in the same cage, so the longest run would be around 3 feet.

Since my initial post I uncorvered some more data.  There is a second Cisco ASA, model 5510, with 4 ports.  Two ports are operating at 1G, and the other two ports are 100Mb.  I believe on this unit the 2 100M ports cannot be changed.  The second 5524 in the stack has ports connected to this device.

I ran a test monitoring the 5524 switch ports connected to this second ASA.  For the test, I copied a 5GB file from one VM to another.  My expectation would be that no traffic should be passing over these 5524 ports as the VMs, by way of the esx hosts, are connected to the 8024 switches, and then to the SAN through the 8024 switches as well.  There is no reason, assuming everything is operating properly, for this traffic to ever leave the 8024 switches.

Prior to the test, the utilization on the 1G ports was at 4%, and on the 100M ports it was 40%.  During the file copy, I saw the utilization on 1G ports jump to 10%, on the 100M ports it increased to 99%.  File transfer rates were at 5-7MB/s, which equates to roughly 60Mbps and correlates nicely with the jump from 40% to almost 100% utilization on the 100M ports.

Again, what I cannot understand is why those ports would show any increase at all.  Another interesting thing is that the 5524 switch ports are all showing heavy transmit traffic, but nearly nothing on the receive side...so it seems it's just broadcasting all iSCSI traffic through the network and the 100M ports are receiving it and limiting throughput.

Thanks again for your help on this!

Eric

No Events found!

Top