Network problems

I really don't think that the problems is on Cisco switches, because some ESX server work fine during the change, but other not. Even I have another c7000 Enclusure without virtualization (the configuration of the Cisco switches are the same in both case), and nothing happened after the change. I even came back to the previous network configuration but nothing work either for this server with problems. One of my question is that all the necessary for restart network service for ESX server are the commands "service network restart" and "service mgmt-vmware restart", I think not, because with the same configuration I reboot the server and everything work fine again . The other question is about the VMotion, why when I WMotion a VM the network stop working until come back to the original server.
Sorry for my poor English.
Best Regard

I send a diagram to facilitate understanding of the network configuration.

Nice Diagram!

a question about it though, I have the same enclosures with Pass-thru ethernet to a 6500, are you using the pass-thru ? If you are using the HP integrated switches, I belive they may connect on the backplane making a STP loop, which could give you similar symptoms to what you are describing

sorry to add to my above:

do
sh spanning-tree detail

find your ports (lots of output sorry) and note the values for

   Number of transitions to forwarding state: X

Do you show a lot of transitions ? Are they incrementing? this can show a STP problem.
 

Check the simple things:

On ESX vswitches that are attached to your physical nic ports (which are aggregated via LACP on the switch side) make sure you're using IP HASH on the vswitches, otherwise this may cause funny things to happen

If I can ask a question on this too ... does STP not have to be disabled on the ports connecting to the ESX Servers? We have to disable STP for all our MLT links going to our ESX servers. We use Nortel Switches..

VMotion wise, I'd suggest you consider separating your service console and VMotion activities from other Virtual Machine network activity.
By this I mean putting the Service Console and vMotion into their own dedicated interface and VLan which ideally would not be part of your trunk or etherchanel group..

 
Thanks for answering.
Hi cat6509,
I do "sh spanning-tree detail" this is the output for the vlan of the Service Console:
Port 838 (Port-channel6) of VLAN0270 is forwarding
 Port path cost 3, Port priority 128, Port Identifier 128.838.
Designated root has priority 32768, address 0011.5df5.190e
Designated bridge has priority 32768, address 0011.5df5.190e
Designated port id is 128.838, designated path cost 0
Timers: message age 0, forward delay 0, hold 0
Number of transitions to forwarding state: 1
The port is in the portfast mode by portfast trunk configuration
Link type is point-to-point by default
Bpdu guard is enabled
BPDU: sent 37903, received 0
This is the output of the vlan for the Service Console, in other vlan the "Number of transitions to forwarding state: 1" are the same. I'm pretty sure that the problem is not in Cisco. The configuration of the Port Channel is:
interface Port-channel6
 description "Encloser Konoha"(2/13-14)
 switchport
 switchport trunk encapsulation dot1q
 switchport mode trunk
 spanning-tree portfast trunk
 spanning-tree bpduguard enable  
The HP Virtual Connect recognizes the Link Aggregation pretty fine (Active/Active), and is configured to work in pass-through with the Enable VLAN Tunneling enable.  
Hi za_mkh,  
As far as I know the HP Virtual Connect does not participate in Data Center STP, neither the ESX server.  
Hi paulsolov,  
Until now I always use the Load Balancing Routed based on source MAC address, I try with this now.  
 
Looking for an answer I see the following in each server with problems. One of the two has a correct range Observed IP ranges but the other card sometimes not see any range at all. Im trying to figure it out why.  
Thanks for everything; I hope that we keep talking.  

The route based on mac address will give you problems in a etherchannel/lacp configuration.  The IP hash should be used.

Thanks paulsolov, I try your recommendation for now on.
Again about the "Observer IP range", something really estrange is happens. When I look for the range of certain NICs they are wrong. For example my IP class is B, and some VLAN has IP range in class C in my ESX hosts, this IP address does not exists in company. I have a doubt, I understand that the virtual switch know the MAC-address of all the VM connected to it. But how does the vswitch know the VLAN??? There is a command to erase this IP table??? Obviously the problem is that. When I try a VMotion to an ESX host with bad IP range in a NIC the VMotion is successful but I lost the network connection, so when I try with ESX server where the IP range in the Active NIC are right everything is ok.
To see the range what I talking about, follow this step:
1.      Click on host.
2.      Click on configuration, networking, Properties.
3.      Click on the tab Network Adapter.
4.      The last field Observed IP range.
Best Regard.

I solved the problem