virtual distributed switches on a single host?

[russ79]

I have two 'datacenters'. One site has several host servers and the other has only one. I created distributed switches at both sites, however, at the site with the single host, it seems to work fine until i reboot. Omce rebooted the host restarts but then can no longer ping or communicate with the network...  The first time this happened, i created a standard switch, moved the management to it and got it back into vcenter where i migrated the interface back to the distributed switch. All was fine until i had to reboot again and once more it stopped communicating with vcenter.... any ideas?

[Virtualinfra]

Is that your vCenter is a virtual machine in problematic site with 1 host.

Then  that is the reason.. as VDS is managed by virtual center. if the host goes down automatically vcenter will not able to manage.

[russ79]

actually my viritual center is located in the datacenter with multiple hosts, so it does not go down when this host is rebooted

[arturka]

Hi

That's it's really odd behavior, never happen in my environments maybe because I always create mgmt traffic on standard switches not on vDS. In case of problems with vCenter or vCenter DB you will  be able to manage ESX networking via vSphere clinet. Does your vCenter DB runs on this same virtual machine where vCenter instance runs ?

Cheers
Artur

Please, don't forget the awarding points for "helpful" and/or "correct" answers.

[russ79]

yes tell me about it... i think ill get it online one more time, move the VMs off and do a reinstall... it makes no sense to me... my vCenter and its SQL DB VM run on a different cluster... were this ESXi 4.x I'd probably opt for a hybrid setup... mgmt on vswitch, everything else on vDS but 5 makes it easy to revert back to a standard vswitch if mgmt connectivity is lost... i just dont get why, after a reboot, it fails to reconnect... almost as if vDS changes are lost on reboot.

[arturka]

my vCenter and its SQL DB VM run on a different cluster.

hi

you should keep your vCenter server and vcenter DB on this same cluster and even on this same node  set up affinity rule and in DRS to keep both VM's together. I will tell you why.

For example - you just added\ modified\ something on vDS and 5 seconds after you reverted back these changes, vCenter replicated changes across cluster (one node by one) data exchanges between esx nodes and vCenter and vCenter DB, if during that "exchange" data process upgrade host with a vCenter DB went down or network connectivity to cluster where vCenter DB was lost you can get into big trouble with network misconfiguration. That's why you should mitigate that risk and keep both VM's on this same node.

Cheers
Artur

Please, don't forget the awarding points for "helpful" and/or "correct" answers.

[russ79]

i think you misunderstood... i have 2 datacenters, one has several host servers, one has a single host... the datacenter with several hosts is where my vcenter and its sql server are tied to a the first host in that cluster, in my second datacenter, i have the one host that is having issues... after some messing around i think it may be switch related... it seems that pulling one of its two uplinks causes the server to start communicating again... now to figure out why