09-09-2013 12:24 PM
Historically we have always installed esxi on rackmount servers using multiple copper nics. We are in the process of setting up our first esxi hosts on c class blades and may or may not be running into virtual connect issues :)
The problem could be with our virtual switch in esxi, with our HP Virtual Connect configuration or with our Cisco Nexus switch configuration. Is there a good tool that we can use on the esxi side that will help us isolate the problem point. The networking guys are blaming the server hardware guys who are in turn blaming the vmware guys and nobody seems to have a clue how to troubleshoot stuff like this.
09-11-2013 12:45 PM
Too many layers to be able to troubleshoot all at once.
If you can ping from Host A to Host B in the same Blade Enclosure (they need to be on the same Network within their VC profiles) then you at least know between ESXi and VC is configured somewhat right.
Then if you cannot ping from a host to its gateway (usually on the Nexus side), then it could be an uplink problem.
Without more info on how you are configured, its hard to troubleshoot.
Here is a generic guide to VC and Cisco integration. However its a bit old so its focused on Catalyst. Your network guys should be able to translate though.
If you are not shy about sharing, SSH into your VC module (same IP, login and pass you use for the Web GUI) and type SHOW CONFIG and save the results to a text file. Then grab a screenshot of the vSwitch screen in VMware. And if the network guys are willing, grab the config from the individual ports and/or port channels used to connect to Virtual Connect.
If you would rather not share openly, you can email me at: dan.robinson
I work for HP so the rest should be obvious :)
PS: Any chance you are in Omaha or Atlanta?
10-01-2013 11:58 PM - edited 10-02-2013 12:07 AM
Thanks for sharing the information here, I wonder if there is a documentation in which I can learn about how is the best to understand VirtualConnect module and the HP Blades c7000 enclosures ?
Because I remember that there are some gotchas or pitfall in managing the Virtual Connect module that I need to avoid like:
- When adding new VLAN in the network for the underlying VMware Virtual Machine, there will be downtime in the blade due to the Virtual Connect must update the Ethernet Network profile in the affected blades.
- Upgrading the firmware of the VC must be done one step higher and cannot be skipped straight away (like current version 3.18 into the latest v4.10 is just asking for trouble in losing the VC domain and server profiles).
- Reconfiguring the IP address of the Blades from one class to another means deleting the server profiles in the whole blade enclosure which cannot be done blade by blade to maintain uptime of the underlying VM.
- ....some more that I don't know of yet probably..?