Force10 and vSphere vDS Interoperability Issue

Recently I had an opportunity to work with Dell FX2 platform from the design and delivery point of view. I was deploying a FX2s chassis with FC630 blades and FN410S 10Gb I/O aggregators.

I ran into an interesting interoperability glitch between Force10 and vSphere distributed switch when using LLDP. LLDP is an equivalent of Cisco CDP, but is an open standard. And it allows vSphere administrators to determine which physical switch port a given vSphere distributed switch uplink is connected to. If you enable both Listen and Advertise modes, network administrators can get similar visibility, but from the physical switch side.

In my scenario, when LLDP was enabled on a vSphere distributed switch, uplinks on all ESXi hosts started disconnecting and connecting back intermittently, with log errors similar to this:

Lost uplink redundancy on DVPorts: “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”, “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”, “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”, “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”. Physical NIC vmnic1 is down.

Network connectivity restored on DVPorts: “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”, “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”. Physical NIC vmnic1 is up

Uplink redundancy restored on DVPorts: “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”, “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”, “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”, “1549/03 4b 0b 50 22 3f d7 8f-28 3c ff dd a4 76 26 15”. Physical NIC vmnic1 is up

Issue Troubleshooting

FX2 I/O aggregator logs were reviewed for potential errors and the following log entries were found:

%STKUNIT0-M:CP %DIFFSERV-5-DSM_DCBX_PFC_PARAMETERS_MISMATCH: PFC Parameters MISMATCH on interface: Te 0/2

%STKUNIT0-M:CP %IFMGR-5-OSTATE_DN: Changed interface state to down: Te 0/2

%STKUNIT0-M:CP %IFMGR-5-OSTATE_UP: Changed interface state to up: Te 0/2

This clearly looks like some DCB negotiation issue between Force10 and the vSphere distributed switch.

Root Cause

Priority Flow Control (PFC) is one of the protocols from the Data Center Bridging (DCB) family. DCB was purposely built for converged network environments where you use 10Gb links for both Ethernet and FC traffic in the form of FCoE. In such scenario, PFC can pause Ethernet frames when FC is not having enough bandwidth and that way prioritise the latency sensitive storage traffic.

In my case NIC ports on Qlogic 57840 adaptors were used for 10Gb Ethernet and iSCSI and not FCoE (which is very uncommon unless you’re using Cisco UCS blade chassis). So the question is, why Force10 switches were trying to negotiate FCoE? And what did it have to do with enabling LLDP on the vDS?

The answer is simple. LLDP not only advertises the port numbers, but also the port capabilities. Data Center Bridging Exchange Protocol (DCBX) uses LLDP when conveying capabilities and configuration of FCoE features between neighbours. This is why enabling LLDP on the vDS triggered this. When Force10 switches determined that vDS uplinks were CNA adaptors (which was in fact true, I was just not using FCoE) it started to negotiate FCoE using DCBX. Which didn’t really go well.

Solution

The easiest solution to this problem is to disable DCB on the Force10 switches using the following command:

# conf t
# no dcb enable

Alternatively you can try and disable FCoE from the ESXi end by using the following commands from the host CLI:

# esxcli fcoe nic list
# esxcli fcoe nic disable -n vmnic0

Once FCoE has been disabled on all NICs, run the following command and you should get an empty list:

# esxcli fcoe adapter list

Conclusion

It is still not clear why PFC mismatch would cause vDS uplinks to start flapping. If switch cannot establish a FCoE connection it should just ignore it. Doesn’t seem to be the case on Force10. So if you run into a similar issue, simply disable DCB on the switches and it should fix it.

Advertisement

Monitoring ESX Storage Queues

Queue Limits

I/O data goes through several storage queues on its way to disk drives. VMware is responsible for VM queue, LUN queue and HBA queue. VM and LUN queues are usually equal to 32 operations. It means that each ESX host at any moment can have no more than 32 active operations to a LUN. Same is true for VMs. Each VM can have as many as 32 active operations to a datastore. And if multiple VMs share the same datastore, their combined I/O flow can’t go over the 32 operations limit (per LUN queue for QLogic HBAs has been increased from 32 to 64 operations in vSphere 5). HBA queue size is much bigger and can hold several thousand operations (4096 for QLogic, however I can see in my config that driver is configured with 1014 operations).

Queue Monitoring

You can monitor storage queues of ESX host from the console. Run “esxtop”, press “d” to view disk adapter stats, then press “f” to open fields selection and add Queue Stats by pressing “d”.

AQLEN column will show the queue depth of the storage adapter. CMDS/s is the real-time number of IOPS. DAVG is the latency which comes from the frame traversing through the “driver – HBA – fabric – array SP” path and should be less than 20ms. Otherwise it means that storage is not coping. KAVG shows the time which operation spent in hypervisor kernel queue and should be less than 2ms.

Press “u” to see disk device statistics. Press “f” to open the add or remove fields dialog and select Queue Stats “f”. Here you’ll see a number of active (ACTV) and queue (QUED) operations per LUN.  %USD is the queue load. If you’re hitting 100 in %USD and see operations under QUED column, then again it means that your storage cannot manage the load an you need to redistribute your workload between spindles.

Some useful documents:

• Interpreting esxtop Statistics (forum post)
• Troubleshooting Storage Performance (VMworld presentation)
• VMware I/O queues, “micro-bursting”, and multipathing (blog post)