stacked powerconnect 5524 address table empty

Question

We believe we are having an issue with our Powerconnect switches.  2 5524 units stacked, connected with two HDMI cables. These are connected to two 8024F switches via SPF ports. Copper connections on 5524's go to VMWare hosts - the 1G copper is solely used for backups from the VMWare environment. The two 8024F switches are connected to the hosts via fiber, and to two Equalogic SANs. Diagram of our layout and the issue is here  (Link to diagram):   Per the diagram above, here is what I have witnessed: Copying a file from one local drive to another on a VM generates traffic on the 1G connections, almost as if the 8024F switches don't know to send them directly to the SANs and they're being broadcast as if it were a hub.  All of the traffic in this operation *should* occur over the yellow connections in the diagram above. This morning I was copying a 6GB file - Windows started out estimating the copy to take over 10 minutes.  About 7 minutes after the copy started, I noticed the throughput go way up and the remaining 2GB copied in about 1 minute.  So it's as if the switches finally found the correct path. I was monitoring traffic in both Solarwinds and vSphere, and noticed receive traffic spike on *all* the copper-connected NICs - even on *different* hosts.  Once the copy sped up, I saw traffic spike on one of the fiber NICs that is connected to the 8024F switches.  In another test I captured traffic on one of the VMWare host's 1G connections then ran through wireshark - it was showing VLAN 68 traffic (which is iSCSI).  Why would iSCSI traffic ever make it over to the 1G side of the diagram above? STP is enabled on both the 5524s and 8024s.  Topology change events happen rather frequently on the 8024s - maybe every 30-45 minutes (is this ok?).  The 5524s report that a topology change has not happened in over 19 days.  Our network is very static - we hardly ever spin up new VMs or remove devices so I would not expect topology changes to occur often. On the 5524 stack, I noticed that the dynamic address table is nearly empty; only 4 entries for one VLAN.  The 8024 address table has 116 entries. My thinking is that the address table being empty might be causing issues with the topology changes on the 8024 side...  Thanks for any help anyone can provide! Eric

DELL-Josh Cr · Answer

Hi, The topology changes seem to be really fast, it shouldn’t change that often. Is the firmware up to date? Are the switches under warranty?

eric_y_admin · Answer

Thanks for the reply Josh.

We performed a reload on the 5524 switches Sunday night and a few minutes after the reload we saw the address table populate with over 100 entries. Performed some tests and the throughput on file copies was drastically improved...and no traffic appeared on the backup interfaces.

We still see STP topology changes on the 8024 switches. Every time I log in the Global Settings page indicates the last topology change was within the hour. Not sure how frequently this should happen in a fairly static network?

Topology on the 5524 has not changed since the reload occurred.

Thanks again

Networking General

Was this post helpful?