Hi all,
I'm configuring smarts to discover VMware ESX hosts and VM running on them. VMware hosts are discovered as standard linux hosts with some problems.
* All vmnics interfaces are marked disconnected and I must unmanage them to avoid seeing wrong major notifications. The only connected discovered interface is the virtual vswif0 (that in my case is composed of 4 vmnics). As a turnaround, using user-defined-connection.conf, I manually connected VMware vswif0 interface to one of the ports in the switch where vmnics are connected but it's not a so beauty solution.
* I'd glad to see a relation between ESX Host and VMs running on it. When I discover a VM it remains physically connected to nothing. I think an ESX server something like a switch connected to VMs and I think it is important to see where VM are running.
If you have VMware hosts in your datacenter how did you managed them with smarts? any advice?
Have you figured out a way to deal with the ESX server modelling in Smarts ?
I am facing the same problem here, I have also talked to EMC about this, they said, EMC Smarts engineering is in the progress to certify Cisco Nexus 1000v virtual swtich, ( not standard virtual switch or Distributed virtual switch), but they are currently not able to confirm the detailed features.
If you have got any idea, would you like to share with me ?
A better suggestion for Vmware implimentation with SMARTS would be ESM( Smarts Enhanced Server Manager).ESM discovers VMware ESX Servers and Virtual Machines, and continuously monitors their availability and performance — automatically.I have also attached the pdf file related to ESM for further clarification.Hope this would help .
Yes, we are also using ESM to monitor vmware ESX servers, but the problem comes to be the physical links between ESX host and Neighbor switches.
From the map console in Smarts, you are only able to see the cable created for vswif of ESX host and switch, but you could not see the link between a
virtual switch ( a trunked uplink port links to a switch ) and neighbor switch, as there is no link between the trunk port on the ESX side and trunk port on
Swich side, we will have risk to lose any of the links.
I understand that, except vswif in ESX, all the other nics are actually vmnics and even in neighbor switch, you can't even find the mac address of these
vmnics' mac address. As far as I know, EMC Smarts is doing some work to certify Cisco Nexus 1000v Virtual switch and I hope this will be make the links
Related to the bulding of the physical connection between the ESX host and the neighbouring switches .I would like to share with you some views made by our experts.Hope this would give some idea.
"Typically an ESX server would not have a single physical connection to the outside world. A bonded set of network adapters is almost universally used. This means that unless you have SNMP enabled on the box you really don't have a good view of the system. However, this also causes other problems since the multiple network adapters typically cause problems with the data from the bridge probe on the switch side"
.
The best way to build the connection is to have them turn on CDP for each Vswitch that is connected to an adapter using this “esxcfg-vswitch -B both vSwitch0” by default the vswitch is in listen mode. Now the esx system will broadcast CDP to the cisco switch an example is shown below
Device ID Local Intrfce Holdtme Capability Platform Port ID ems-labesx02.isus.emc.com Gig 1/0/2 127 S VMware ES vmnic1
Regarding the official certification of the Nexus switches are in progress as previuosly suggested by you.If you need further clarification I would suggest you to open an SR with the support team.
Adding to the previous post regarding Nexus 1000V switch certification, I have the information that this switch is officially certified in the latest patch ( patch 28) for IP 7.0.3.You can have it downloaded from powerlink.
Happy to hear that you have already tested the certification for the Nexus 1000V switch. I will also let you know in case I have any further helpful information related to this.
Just want to let you know that, finally I got the CDP post discovery working in my testing environment.
As I said, I opened a case with EMC support , they investigated this issue then went back to me that, the reason EMC Smarts IP-AM could not create the connections during CDP discovery is caused by the different CDP message broadcasted by virtual switch/Distributed Virtual Switch. [ different with the standard Cisco CDP message). They did some modifications in ic-cisco-post-processor.asl for me and tested successfully in their env, but unfortunately, the one they modified still doesn't work me. After some debugging, I noticed that, that is caused by the multiple bridges discoverted in the same port ( there are a few vms' traffic cross that trunk port). I addd a few lines to get ESX servers treated differently then sort this issue out.
According to EMC support, EMC probably will release a new ic-cisco-post-processor.asl which will fix some of the similar CDP discovery problems, maybe in next IP patch. But I am assuming, we still need to do some tweaking to make it working for ESX servers.
I am very happy with what EMC support provided me.
-------
For the Nexus 1000v, I made it running in our lab, from what I can see so far, it is treated as an individual Switch in Smarts, and the links to physical world can be created without any problem. But I am unable to see the links inside the ESX hosts, which are the links between VMs and Nexus 1000v Switch. I was expecting to see this, some of the customers are eager to see this as well. I am not sure if I did something wrong on the configurations or there is a mib issue for Cisco Nexus 1000v. I will do some more testings when I get a chance.
I just want to check with you, do you know, with Cisco Nexus1000v, are we able to see the links inside the ESX ( between vms and Nexus Switch) in Smarts.
I have done some testing in our lab, but I still could not get the links up. I can see the vm nic’s mac apprears in the mac table in Cisco Nexus 1000v by suing "show mac address-table", but for some reason, in the sm_snmpwalk file, the mac address table keeps changing. The changing I mean here is: the mac address of the vm’s nic sometimes it is there, sometimes it is not there. So from my understanding, it could be a mib issue, or there is something with the configurations I've done in Cisco Nexus 1000v.
hubery
13 Posts
2283
0
Posted July 5th, 2010 17:00
Hi CeSIA,
Have you figured out a way to deal with the ESX server modelling in Smarts ?
I am facing the same problem here, I have also talked to EMC about this, they said, EMC Smarts engineering is in the progress to certify Cisco Nexus 1000v virtual swtich, ( not standard virtual switch or Distributed virtual switch), but they are currently not able to confirm the detailed features.
If you have got any idea, would you like to share with me ?
Thanks,
Hubery
Message was edited by: huberyhe